BsidesLV 2024 - Breaking Ground - Wednesday
Show transcript [en]
[Music]
I oh
[Music] [Applause] [Music] [Applause] [Music] oh
[Music] [Applause] [Music] I'm just tring to I'm just try to give you [Music] something I'm just tring to give you something I do I'm just trying to give you something [Music] oh [Music] [Applause] [Music] [Music]
[Music] [Music] I'm just to I'm just TR to give you [Music] something I'm just TR to give you [Music] something I'm just trying to give you something [Music] m [Music] w
[Music]
[Music] [Music]
[Music] is
[Music]
[Music] [Applause]
l oh [Music]
[Music] [Music]
[Applause]
[Music]
he la [Music]
n [Music]
[Music]
[Music] [Music]
[Music] [Applause] [Music]
[Music]
[Music]
[Music]
[Music] [Music]
[Music] [Applause] [Music]
[Music] a [Music]
[Music]
[Music] [Applause] [Music] hey hey hey he he [Music] [Applause] [Music] he [Music]
oh [Music]
[Music]
[Music] TR [Music] hey [Music] [Applause] [Music]
hey hey hey hey hey hey hey now [Music]
[Music]
[Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Music] [Music]
[Music]
[Music] [Applause] [Music] he
[Music]
[Music]
he
[Music] h
[Music]
[Music] [Applause] w w [Music] [Applause] [Music]
I'm I'm just try to give you [Music] something I'm just try to give something I do I'm just tring to give something there [Music] [Applause] [Music] [Music]
[Music] [Music] I'm just to give you something I do I'm just TR to give you [Music] something I'm just so okay I do don't I'm just trying to give you something [Music] m [Music]
[Music]
[Music] [Music]
[Music]
he
[Music]
[Music] [Applause] oh
[Music]
[Music] [Music]
[Applause]
oh
[Music]
[Music] a [Music] a
[Music] oh [Music]
[Music]
[Music] [Music]
[Music]
[Music]
[Music] a [Music]
[Music]
[Music] he [Music] [Music] [Music] [Applause] [Music]
[Music]
[Music]
[Music] a [Music]
[Music] [Applause] [Music] he [Applause] [Music] [Applause] [Music]
[Music] he [Music] he
[Music]
[Music]
[Music]
[Music] track [Music] hey hey hey hey [Applause] [Music]
he hey hey hey hey hey [Applause] [Music] he [Music]
he
[Music] h
[Music]
[Music] now [Music] [Applause] [Music] [Applause] [Music] oh
[Music] I'm something I'm just TR to [Music] something I'm just try to give you something I I'm just try to give you something [Music] a [Music] a [Music] [Applause]
[Music]
[Music] [Music] I'm just I'm just [Music] something I'm just [Music] something I'm just trying to give you something oh [Music] w
[Music]
a
[Music]
[Music] [Music]
[Music]
[Music]
[Music] [Applause]
[Music] [Applause]
[Music]
[Applause]
[Music]
[Music]
[Music] n [Music] oh [Music]
[Music]
[Music] [Music] [Music]
[Music] good morning day two besides Las Vegas congratulations to all of you for making it back into the conference after last night uh I hope everybody's remembering to drink their water although you're probably not the ones I have to talk to about that but keep drinking all day hopefully some water mixed in there um we have uh a couple of uh minor announcements so uh reminder to everybody so we've got the uh the sky talks lines this year which are new right um off in the the corridor on the side of the uh breaking ground here uh the lines are long but they move really quickly once you know once they open from the uh the moment
that the uh the tokens are distributed for the next sessions um if you're not familiar with that we have Sky talks over in the Platinum building this year up on the the penthouse and uh you can get a token here that allows you access to a particular session there's uh series of them through the day and there'll be a variable number of talks within a particular session so if you look on the schedule every one of those talks tells you what session you would need access to you can go and get in line over here when the um excuse me the tokens for that session are being released if you're interested um excuse me all right that
said uh I want to thank each and every one of you for continuing to mask uh and I know it's it's not fun but uh we really appreciate what you're doing for the folks who need that and uh I think that's it is in terms of things that we need everybody to be aware of so uh while our esteemed uh guest is getting ready I'll talk a little about this is we have with us today um Professor Andrea matsan from uh University of Pennsylvania and Penn State Penn State excuse me I yep I used to teach it you so you're not wrong it's it's somewhere in that Eastern pen sort of I've done I've done Northeastern
Northwestern pen Penn State uh she she has a very uh illustrious uh set of credentials but she's also a a professor uh both of engineering and of Law and has spent a lot of her career looking at the intersection between the two and the policy between the two uh and and how we can be more effective in how we think about regulating and um you know governing our use of Technology uh this is all as you might imagine very relevant in in light of some of the things that we've seen in the last uh course of this year uh around uh saf and uh uh serious you know human and financial impacts resulting from our uh
use of Technology uh so she's going to be talking a bit about how we can take a look with a through a a a safety lens uh sort of the the traditional policy Frameworks that we've used in areas where life and and you know critical safety are important and uh what that might look like so that we could have a a a more effective next couple of decades than what we've seen in the last you know 20 years of of technology and and uh some of the outcomes that we've seen there for safe Safety and Security so uh I'm going to stop talking about it we'll give her a coup know couple of minutes while they're working on the because of
course in the live version yes it's not going to work the way it did but uh you know I think something very relevant for uh the moment and and uh something that could have some significant impact uh and start a very good conversation about how we should be thinking about these incidents and how we get out of uh the game that we have now uh around uh underinvestment in security and safety and excuse me sort of externalization of uh some of the risks that are involved there so uh in with that I'll uh turn off the mic for a second and uh we'll get going as soon as this uh AV stuff is sorted out
hey worked find a
test there we go we're good to go yes Mr Corman I'm going to have to ask you to take your seat please thank you all right without further Ado The Fabulous Andrea mishan not sure about the fabulous but thank you very much the check is in the mail as long as you're sure as long as you're sure about the Andrea mition part we're okay most days okay can't be 100% most days yeah so thank you very much for the invitation to speak today thank you all for showing up at 9:30 at a hacker con morning that is hard I know um so uh I will share with you some thoughts um some of them will be
intentionally a little bit provocative hopefully and I look forward to engaging with all of you on the topics that I mention in my slides also I have quite a few slides so if we get to the 45 minute point and I'm still doing this someone Josh Audi someone please pick up the St uh du uh uh hand thank you okay so this talk is titled homicide wear um for reasons that are probably uh somewhat obvious to this crowd because increasingly we are starting to be worried about the Confluence of lifethreatening events that endanger both life itself and physical safety of human bodies so before I launch on my thoughts on this topic just to give you a brief sense of who I am
where I've been I've been doing this a while um and for better or worse it is my 25th Law School reunion this year not sure how I feel about that I'm speaking at it also not sure how I'm feeling about that one um so I've done some government stuff been in this community it's been a great pleasure since 2003 um and I look at things maybe a little differently because my background is a little different than um most of you in this community My Views are mine please do not attribute anything that I say to the nice people in government that I work with I'm not their fault I'm me um also this is original academic research
so please discuss this to your heart's content but please also do remember where you heard it because it's kind of Soul crushing to have you know 20 years worth of work summarized in a one paragraph blog post that may or may not be quite right without attribution um so please thank you okay away we go in case you need to take a little nap at any point during this talk I'm going to summarize everything on two slides security is the Lynch pin of Technology safety but we have generally not been talking about it this way in policy and legal circles we have instead kind of essentialized exceptional lied the questions of Technology as they relate
to issues of physical safety but also economic safety emotional safety Financial safety all of these are bound up in the issues that this room thinks about regularly but the policy and legal discourse conversation hasn't really mapped well so confidentiality Integrity availability issues are regularly now threatening public safety questions both in terms of infrastructure but also in an individual level but here's what's shifting now and I've very um kind of mixed feelings on this one the normies are starting to get scared the kinds of questions that I'm getting from my non-te friends now are shifting that include and they include things like oh this is what you've been trying to warn me about for the last 25 years so that
is a positive Evolution but it also means that they're having that initial um moment of trepidation fear that some of us uh May have experienced when we first realized that all we've got to defend us is us so the way that we tell the story to policy folks to normies to ourselves about the last 25 years of what we've been doing in security is going to shape the next 25 years of security this is a great point to pause do a little bit of self-reflection on the way that we've been talking about these questions and whe whether the way that we discuss them in community are the best way to talk about them with people outside the
community and figuring out how those things fit together so I'm calling this conversation a Goldilocks problems situation which you'll see in a minute uh hopefully with a bit of humor but there are some hard truths that I think we haven't quite discussed amply about Insider attacks the uh perhaps inadequate threat modeling for Insider attacks the internal controls inadequacies that are rampant inside our companies that are very sophisticated in some ways and kind of let stuff slide through the cracks and other ways and the language of internal controls is the legal language so some of these terms that you know what I'm talking about but they look a little weird it's because they're the way that the legal framing
would engage with this as we enter the age of AI security which you had a wonderful keynote about yesterday things are going to be more complicated not necessarily new but more complicated and even more about tech safety for reasons that we'll briefly mention and increasingly the line as we all know between civilian safety questions and national security safety questions is functionally gone because of the revenue models of some of the uh threat actors that we know exist in other parts of the world so the public relations spin about harming innovation in Tech creation it's just not going to hold water going forward it's not going to stop liability because the Frameworks that will be applied are
not those from technology spaces they're from law spaces and the way that legal uh F finders of fact we would say judges juries uh Regulators think about these questions are going to be very harm focused they're going to look at the bad thing and work backwards they're not going to start with the tech question so how should we think about this if you just hang your hat on be it hopefully a white or gray hat um on the idea of context harm and intent knowledge and I will unpack each of those as we talk those are legal terms of art that will help you get a handle on the way that a judge would look at these questions the
way that a regulator will look at these questions so the context that you're operating in the kind of harm that's possible and who knows what and controls what or who should know what and control what and we'll walk through the example of homicide homicide you generally think about first-degree murder you think about the attacker who premeditates something for months and executes it but that's not how the law thinks about homicide there are lots of very spur ofth the- moment situations that are very serious felony level of prosecutions and so homicide with a computer is still homicide and we'll walk through that example to talk through things so I'm not going to leave you with all gloom and doom we'll also
talk about what we can actually do and this is a position that I would not have adopted 25 years ago this is fresh and new to me because of my work in government frankly and seeing the tech economy shift in some ways that um I find concerning and that have um made me fight to remember the joy that I felt opening my first browser and telnet and those bygone days of the green and black um and I think helping to share that history with the youngans would um instill a little bit of Hope in the way they were dealing with these questions and remind ourselves that there is still hope so I think we need a new regulator
I think we need a technology regulator of Last Resort not to change things dramatically just to fill in gaps to coordinate to create alignment as the techies would say um on various matters of policy and law and to help stand up interesting new Gap filling initiatives in the most critical sectors like healthcare where the security problems are not easily handled by small hospitals on a local level they need help um and they need coordinated help so Tech safety is a different framing kinging on context that should be context sorry harm and intent and beware the uh exceptional iing of AI the last point which is just a teaser for the workshop that will happen immediately after this
we'll be talking about professional Society emergence and I know this is anathema but it is a part of every mature industry sector's evolution where you get to reclaim your story and tell it the way you want it told and so this is about storytelling and this is about building the next 25 years in a better way so that we don't feel like each year we're just banging our heads against the wall a little harder so please come to that Workshop immediately after and um there will be stickers so many stickers um so even if you only stay for a little bit come for the stickers okay let's start the Story Once Upon a Time an attack or
compromised a physical perimeter of a physical plant the confidentiality of the information the Integrity of the internal manufacturing processes and the availability of critical infrastructure resources once upon a time a potential data science entrepreneur leveraged Guerilla collection tactics to generate assessment metrics for residential infrastructure supply chain inputs manufacturing processes and Qi QA using trial and error and high reliability low latency sensing technology once upon a time a Lost Child sought refuge in a structure and survived on foraged food until rescue once upon a time a trespasser entered a bear habitat ate the day's rations broke the UR sign enrichment materials damaged the living structures and potentially germed up the den with human pathogens dangerous to the Cub's health
these are all the same story they're all potentially simultaneously true so what does this mean it means that we're living in a complex situation where narratives matter Goldilocks is simultaneously a criminal Intruder a budding data scientist from the Bear's perspective a destroyer of habit and the Bears would view themselves as the center of the story not necessarily the human child this is also a story about the intersection between lawyers hackers and bears they are all dangerous when they're hungry and let me assure you that at this point the lawyers are getting hungry and what will make the future rollout of uh the crowd strike litigation that's already been filed and will be filed um very uh educational
interesting from a legal observer's perspective is to see how that plays out in terms of the claim allegations the settlement rates Etc and the insurance Dynamics which are going to be messy so some facts about context harm and intent are going to be disputed so what do we do with this well in Goldilocks and the Three Bears there are actually many versions of the story where even the main character changes so these kinds of tellings and retellings are something that we are all familiar with in the time we're kids but what does change is whose story it is what the perspective is from which it's told whose progress is centered in the evolution of the
story so in the last 25 years of security there's sort of been a default Assumption of a free ride on the part of many compan of a default of no liability and the conversation in legal and policy circles has been very sort of focus circling around questions of breach remediation and compliance which is a word that I am not a fan of um as my meme there demonstrates um in the public sector we've had a national security Focus that has been I would argue too narrowly focused on the cyber what in this community jokingly we say pew pew right a little bit too much on the Cyber pew pew aspect of it and on the enforcement of confidentiality
failures of systems rather than looking at for example how serious the availability harms and the availability failures can be in terms of Simply the maintenance of everyday life for increasingly millions of people around the world and the rates at which particularly body embedded devices are being adopted did in medical but sort of um medical contexts where there are non-invasive Alternatives available it's something that really gives me pause as to whether we're building more than just additional medical infrastructure so there's really no more bright line between National Security concerns and civilian safety concerns when we're looking at the way that threat actors and Insider attacks can happen and on the ground impact human bodily safety Financial safety infrastructure safety
and the future of our own economy and Country and democracy and this is going to get progressively more problematic in AI context so let's start to think about how we can retell this Story Once Upon a Time very long ago it was 1999 and there was a Y2K crisis that was ending and and this story and if any of you were at my uh Defcon policy talk last year you know that um I'm really bullish on retelling the story of Y2K because I think that the people who were embroiled in it at the time suffered post y 2K traumatic stress trauma and didn't um uh want to talk about it then as much but maybe now enough time has
passed that they can talk about it um there was a successful hole of G government response with hearings and plans and two statutes were passed and public private sector cooperation worked there was actually very little disruption in our infrastructures and in the economy there was litigation anyway there was a lot of litigation but it really could have ended differently and that's the part of the story that's getting lost why 2K is too often a punchline it wasn't a punchline it was a success story and we need to reclaim that story and tell it every time someone says that response doesn't work security isn't worth investing in this is one of those case studies that you
can point to concretely but something else I've been thinking a lot about lately is the first generation of viruses and worms and they weren't motivated by intention to steal but they still had huge availability impacts and so when we look at the way that we thought about these two scenarios in 1999 Maybe by putting ourselves in the time machine a little bit we can reconnect with earlier versions of ourselves and start over reclaim the next 25 years because really I think it's our last chance in terms of making sure that we don't suffer an e economy-wide catastrophic security related event that isn't as easily recoverable as the uh availability incidents that we've been recently dealing with so with the Melissa virus
stuff there's one um prosay case where someone accused a private party of talking smack that he was spreading the Melissa virus no big deal but with the Y2K cases despite the fact that we had two statutes there was substantial litigation but it got resolved and through a lawyer's eyes that's just normal there's going to be litigation people breathe there's litigation that's why you have lawyers but here's what we should think about in particular when you look at the Y2K act even though it was tailored narrowly about a very particular event with a commonality of interest that was narrow and we cannot pick it up and wholesale apply it to our current complex context of security Dynamics we should learn
from the way that they thought through some of these issues and they may give us kernels to chew on and to think about Colonels in the corn sense not in the computer sense to think about how to uh move forward in the way that we talk about our current security issues so they defined material defect they explicitly Exempted any claims for physical injury that's big they created a special contract right and for most of you this may not be particularly interesting but from a lawyer's perspective this is huge this is a when I saw this reminded that myself that this existed it was a little bit of a mind-blown situation we can talk about that in contract nerd study
hall later um they were conscious of small business concerns they were conscious of the Dynamics of various control entities they were conscious to take out Securities litigation and enforcement by other agencies and here's the one that I think May resonate most with this community they developed a process for including true expertise for the most complicated cases so there was a way for a judge to appoint a special Master to appoint someone to help unpack the toughest questions so here are the lessons context really matters if you focus on each of the issues that's understood and approachable you can generate cooperation and you can build a process that includes clud expert input harm comes often from disruptions
and there will be potentially deadly effects unless mitigated and so the key question that we're facing now is still are the the problems at hand the incidents in hand disrupting the confidentiality integrity and availability of systems and will those impacts be felt by particular groups of people who may be particularly vulnerable and may experience particularly severe harm physical economic Etc and finally intent and knowledge calculus in law is nuanced and we'll we'll unpack this but this is I think the main thing that I hear Executives not understand that intent is not just what you do know or what your lawyers told you and and can be proven with documentation they told you but it's also what in light of your role you
should know and what other factors around you should have informed you if you were paying attention that is the best course of conduct that a reasonable person in your situation would engage in all right so back to that talk last year I kind of uh set out the the warning Beacon that liabili already here liabil coming that non- liabilities actually I would argue a form of legal Tech debt that we can't keep writing on and that it's going to take various forms private litigant civil regulatory action and potentially criminal harm and then I wrote this post in January about boot loops and security and I made some proposals including the one about the bureau technology safety and um the
importance of messaging the importance of having those interventions the importance of a new agency and I did not anticipate that uh we would be facing an availability uh incident uh that penetrated the consciousness of everyday folks so increasingly when we talk about security I would encourage us to think about security as the lynchpin of Technology safety and that the whole Enterprise is about keeping people and things safe that is a framing that resonates with policy makers and with everyday folks it will resonate with your grandma it will resonate with your aunt it will resonate with your 9-year-old and the best test for whether we're messaging well and whether we're talking about our own life choices and
careers well is whether we can explain what we're doing to a 9-year-old they are a very tough audience as someone who just spent brunch with a 9-year-old a couple weeks ago earning her attention was a trick so tell telling things in straightforward ways that make sense is um a skill that um I think it would be important for certainly myself but I think the more we can all do it it's helpful so the push back that I always hear is oh but what if it's just a defect what if it's not a malicious attack what if it's just an Insider error context harm and intent will capture all of that because ultimately the focus is still on the potential harm
it's still a safety incident even if it was an accidental safety incident so for example the one that's top of mind we're all familiar with the recent incident that uh resulted from um an update that was pushed um by crowd strike um and so $5.4 billion of damage is one estimate 25% of the Fortune 500 disrupted according to this estimate um and 100% of Transportation sector disrupted allegedly according to this estimate and the banking impact was one that particularly concerned me but we also had 911 systems disrupted we had hospitals disrupted flights disrupted we had again with the banks disruption um and this lasted for um most of the day at at least in some places
um and this was a blood supply problem for example in the case of New York blood center so uh assuming all this reporting is correct there were many different kinds of entities that were impacted and the mayor of Portland announced a Citywide state of emergency that's bad in terms of potential impact when a city feels then this is not a debatable technical thing this is how the city feels about its own State of Affairs the city felt that it was in a state of emergency when normies feel they are in a state of emergency that's definitely a safety issue regardless of how we frame the technical discussion that underpins that safety issue and indeed um sisa um and kudos to
them released um a set of uh reports and updates about the incident including a post incident review but we also saw that other kinds of uh unexpected followon activity was visible including some scammers using this as an opportunity to push out new forms of malware and to exploit this uh public concern in ways that were advantageous to them as criminals and frauders and again damaging to safety of the public so of course the social media conversation in this community was um aggressive shall we say and this m pointed out the experienced reality of these availability incidents and this is how normies perceived it too they didn't they didn't necessarily understand what was going on at first um they were
worried there was was some sort of large scale attack um and it got to the point where um AARP was sending out emergency bulletins to the seniors to tell them it's it's okay it's going to be okay right so when you have panicked senior citizens it's a safety issue okay so I found this interesting from a legal standpoint the characterization of this update incident as not a security incident but as a defect so defect is a an an interesting legal word there um now of course the commenters you all among them no doubt we're quick to point out that an availability problem is one of the categories of the CIA Tri Triad right so there was debate about this
characterization naturally but there were consequences for the entity that again to a lawyer's eyes immediately trigger the likelihood of certain kinds of litigation whenever you have an 11% Plunge in value uh you start to anticipate the likelihood of suits from shareholders because that is commonly viewed through lawyer eyes as potentially enough above the threshold of what would constitute a material change for Securities laws purposes so you're dealing with a public company there's that layer of Securities Law and likely a corporate law follow on but I'm getting too much into the weights but here's something that I just want to share as a little signal that I saw this headline and I was like oo David boy is a very expensive
lawyer so when I saw that Delta was investing in David boys that was a signal to me that they were very serious about this litigation and so we also have other pieces of the legal ecosystem that are starting to flourish around looking for opportunities with potential plaintiffs so the lawyers have arrived in ways that they hadn't 5 years ago even so here's just a quick sketch and I didn't get a chance to run the search again last night to see what was new but we do in fact have a shareholder suit that's been filed uh in um connection with the crowd strike incident um Delta has of course signaled uh their interest in suing or
settling um and we're likely to see tort meaning civil wrong suits and lots of contract litigation potentially some infrastructure disruption issues potentially some regulatory enforcement and potentially some physical harm suits so um sadly I know of one case personally where a friend's teenager who's um uh addressing um successfully happily addressing lymphoma could not reach the hospital and had a time sensitive medical issue going on so happily that case were out but it makes me think that maybe there's the potential particularly in light of the hospital disruption of some cases that perhaps um will be brought in connection with harm so suits will be filed under like under various theories of Law and this isn't just
about the crowd strike litigation at this point I'm talking generally about this kind of scenario so this is not me trying to harsh on crowd strike um there will be various theories of law that will be filed that was the point of the last Slide the litigation will go on for years and there will be appeals and that will go on for years this is a multi-year thing that will be very very very expensive for everyone and very disruptive to business ounce of prevention pound of cure really that's the thing that just keeps sticking in my in my head um Whenever there is an incident um that involves something that is preventable um and we will see to
what extent that is true here as the cas is progress us now um some claims will be delayed because law world does not move fast on a typical contract suit in most cases under most state law because contract law is state law that means lots of different state laws but generally you get up to two years to file for a contract suit that means there can be a dribble of various litigation happening for quite a while and and some settlements will be paid out to end threats of litigation in these kinds of situations and that's a prudent choice in many cases depending on the particular facts and circumstances the context the harm and the intent knowledge are always going to
be those three magic variables so um sometimes you end up in a renegotiation or a contract termination in the parties go their separate ways and also we might have some regulatory action there is at least one enforcement action from a financial regulator cfpb that focused on availability issues so we might see these kinds of enforcement actions uh get brought in terms of the various Regulators um scope of um Authority under their enabling statutes so um this will be an exciting time that we're entering into in terms of litigation and lawyers so when I talk about context what I mean it varies by place and jurisdiction it varies by the community and how they've defined harm in statutes and
among themselves it varies in the emerging effects that happen sometimes you can't necessarily anticipate fully what the context will hold for you it varies across time laot evolves and so does technology and so do the ways that people interact with them so the question is one of suitability of the design and the internal controls for that context so let's come back to some bears we're all familiar with these hacker Bears uh for better or worse but you might not be familiar with this hacker bear who let himself out of a zoo multiple times and needed to be moved to a different Zoo design with a moat you might not be familiar with actually this one was pretty widely
known but nevertheless you have to respect the bear you might be familiar with this bear who has uh a dedication to pursuit of lasagna who led himself into a Connecticut home open the freezer pulled out the lasagna and left without incident mission accomplished you might be familiar with this bear who just likes to hang out on someone's picnic table in their backyard different context different threat profile different possible harm you might have heard about this Alaska bear that decided to just hang out on a roof until the Skylight collapsed so that he could fortuitously get access to Cupcakes different context you might have heard about this Canadian bear that broke into a car and guzzled 69 cans of pop that's a very
different threat profile on that bear than the one that walked 200 kilometers for Revenge so in Ontario there was a bear that broke into a set of cars not once but three times they took him away the first time he came back and they are sure it was the same bear because they tagged him the first time so this bear walked 200 kilometers for Revenge that is a very different context a very different threat actor and a different response and a kind of Amplified harm because I think they're down four cars now so context matters and the the owner of the cars in this case said normally they're no problem but honestly this one was known
to be bad reputation matters too so harm so I've been doing some research looking at historical catastrophes broadly defined and this is still research in progress but I wanted to preview some of it because as I'm going through this increasingly there's resonance with the types of issues that we think about in security so if you know about a local catastrophe where death resulted in connection with an engineering safety choice or lack of choice and please do share that incident with me because the resources are not complete they are kind of scattered and not necessarily curated in ways that are tailored to this framing but I think this is an interesting framing for stimulating discussion for us so these are
catastrophes where death resulted and there's an engineering Choice design choice that was involved so far I have over 120 and my sample it's going to keep going up um and so basically this is a traditional uh social science methodology I qualitatively coded each of the cases in the sample using um in in my case uh a basically a binary I I yes no question around whether there was an uh an engineering safety problem that arose um leaving in the ones that that did and what through my eyes a court would determine to be the cause the butt for causation the underlying cause they're just legal footnote in in um Tor law in particular in civil wrongs there
two kinds of causation but for cause cause in fact and proximate cause proximate cause varies a lot across jurisdictions and people fight that one out hard but the butt for causation less controversial usually and so what this started to crystallize for me is at least 10 lessons that I think this research might um be useful for stimulating discussion in this community so I kept seeing a repeating pattern of failures to remediate technical debt failures to finish in complete projects we have Bridge collapses with cracking cement partially because of uh things falling through the cracks no pun intended in terms of addressing incomplete parts of the design there are warnings that are ignored repeatedly users testers employees uh even the
actions of neighborhood children sometimes point out that there is is a problem and management has actual knowledge of the problem but chooses not to act for a panoply of reasons there's also a recurring trend of failures to usability test for human operational error so the interactions that are going to happen on the ground and even when there is a human in the loop they're frequently set up to fail in terms of the way that the controls are set up or the way that they are likely to get distracted by something in their environment sometimes you even have fail safes that are in place but they get turned off and so we see train derailments we see oil platform
explosions that fall into these categories um we see a recurring pattern of failing to threat model to anticipate failure scenarios in foreseeable contexts of deployment and in lack of a planned response so grain silos and molasses making operations frequently apparently have exploded um and so if you were choosing to engage in a line of business that has a history of these kinds of events depending on how you design your plants and Technologies um that gives you a uh heightened need for thinking through in your threat modeling how you're going to address those things regular L I'm seeing things that are built too fast and unsafely for financial and profit reasons there's a motivator that's an
external deadline that has nothing to do with engineering quality nothing to do with engineering safety but there is some sort of financial Target of opportunity they're trying to hit and people choose to look the other way on safety to hit the financial Target there's a failure to test appropriate tools and materials so supply chain issues and we see this happening with construction crane collapses regularly which is still a problem um sadly um that we're we're working through design choices sometimes conflict directly with safety and those are sometimes done for uh aesthetic reasons something looks nicer in a certain way um or the person who is making the call is not up to dat on what
is considered best practice at the time so they might use for example in building contexts a kind of plan layout or um uh choice in the build in in basically putting up the building that is faster but is known to have serious safety implications um so that's a design choice there's also a regularity in these incidents of deviations from plans so even when the plans were made correctly in the first place something happens on the fly in the course of deployment that causes someone to cut a corner or to not point out or correct uh a defect in fact as it's happening because it's kind of annoying to fix things and so convenience sometimes leads people to cut Corners
leading to catastrophe building specs um are one type of recurring problem here failures to maintain or respond to incidents adequately Bridge collapses have regularly happened due to neglect and people just forget about checking out if the bridge is still okay because that's a maintenance thing not a splashy headline thing that's a going to the dentist kind of thing but you know we all need the dentist and finally a failure to adjust to emergent changes environmental changes or interactions with the Technologies themselves it's how we had the key bridge problem the barges kept getting bigger and the bridges stayed the same across time and so there was just less and less of a space for uh forgiveness in the driving
all right I will I will uh give you a quick example of a counterintuitive case that um I found particularly interesting 1919 had 21 people die from a molasses explosion in Boston The Operators of the plant blamed the incident on Italian anarchists however locals had observed structural issues with the Molasses tank in fact we know that there was a wall of molasses that was powerful enough to destroy a fire station and it was in January and so 21 people ultimately died some from the initial 50ft wave of molasses which was not slow and then the rest Rescuers got injured and some of The Rescuers died because of these circumstances so there was a supply chain issue in that the
Molasses had been or at least the inputs to the Molasses had been transferred without being allowed to adequately cool the tank itself was known to be too thin in retrospect but even at the time it was not to Speck according to the engineering experts who reviewed this the workers in the plant told their bosses hey we've got a leak the neighborhood children keep coming around to steal molasses from the holes in the vet that's a sign but because this was in connection with a speedy window of production for the War uh era they were so focused on meeting those deadlines that they painted the Vats instead to hide the problems with the cracks so there was was also a legal
angle this ended up in litigation the company was ultimately found liable and one of the quirks was that it had been licensed the Vats had been licensed as receptacles not as the standard full pering process for a building and so this caused a shift in the way that the engineering profession and Architects did their work going forward it's attributable mostly to this particular incident uh in the way that the engineering historians retell it and created a sense of more of a professional duty to safety and independent reporting of of incidents but we haven't solved the Molasses problem fully but we have made it less deadly generally so 2013 we had a pipeline issue in molasses in Hawaii and
there were no deaths in that case but there was no human deaths but there was Mass death of Wildlife and some pollution from it so 100 plus years later we'll still we're still struggling with molasses safety but the interaction of technology today threatens to unwind some of the progress that we've made in the past because those tank issues might end up being over trusted to a piece of technology that itself is flawed and of course as we all know part of the challenge is that it's a two-way street the maintenance is ongoing the maintenance continues and a pushed update can change everything even if you have the original set of circumstances under control so um back to the intent
question because I think this will interest uh you all in this audience there's a slippery line that can't necessarily be placed well at the time of an incident as you all know between an act of an external attacker and potentially an act of an internal sabator or someone who had a bad day so figuring out where the reality of this particular context this particular set of harms and this particular case that is going on at the moment whatever it may be that assessment of what the intent drivers were what people were thinking what they were doing in their hearts it's not going to happen on the Fly you're focused on remediating the immediate issue and the incident
response but retrospectively that's what courts are going to look at they're going to look at who are you what is the role of your business in the economy what is the expected knowledge for someone in that role you chose that line of work you chose that business model you chose those employees whom do you control what do you control what choices did you control and for individual liability questions and there have been some cases recently that I know have been top of mine in this community with personal liability issues you look at yourself your own professional history what knowledge you have been exposed to what knowledge someone similarly situated would have and should have and then you deduce a
third party finder of fact objectively looking at the situation from afar whether that's a jury or a judge or an enforcer says hm based what I see objectively do I think this person made a certain kind of choice and then that determination about the state of mind for that choice guides the determination of what legal consequences follow so in civil contexts you don't have to foresee the full extent of your harm and this is one of the biggest misunderstandings you just have to intend to engage in some conduct whatever harm follows is on you you made the choice to take that step what happens after is attributable to that step so that's why it matters
whether you had knowledge whether you used care and what you promised people in criminal contexts there are specific levels of knowledge that are listed in statutes and it's one of the things that legislators fight about and and lawyers fight about whether you've met that level of knowledge and they're also tort meaning civil corollaries to most of the criminal causes of action so here's as promised a walk through homicide we have 50 plus versions of homicide statutes potentially this is a general intent crime what does that mean it means that the actual intent to do some act as I was just saying is the operative question not necessarily whether you chose the result so there is murder which is by degrees
first second third that varies by jurisdiction what that means exactly and what's entailed and which ones they have there's felony murder which means someone dies in connection with your committing a different felony there is also something called depraved heart indifference which means that you functionally take control over a situation where you know the person is not going to be able to defend themselves and then they die that is a different category that falls into this murder Spectrum there's an intentionality about it but intention can be expressed or implied the magic language of intent in criminal context in this murder context is malice a forethought it's just the way we talk about intent there's some common defenses but there's also a
category of manslaughter which means that you knew you were doing something but the full ex the the full nature of what you were doing was perhaps not as easily anticipated by a reasonable person so there's voluntary there's involuntary there's vehicular there other special kinds and this is all specified by Statute this has come up recently because of Boeing so as we all know there was a sensor issue with the Boeing planes and there was software implicated and the families of the deceased filed wrongful death claims there was a lot of litigation Boeing settled nearly all of it so we don't have a lot of case law on it but we have reporting about just in the one flight
there were 171 people's interests represented in the lawsuits that were filed and 140 of the of 150 claims in the northern district of Illinois were partially or fully settled so that's just one jurisdiction we have a chief executive who admitted under oath that he had knowledge of whistleblower uh complaints and potential retaliation um I watched that testimony the question that I had uh was um why that particular CEO wasn't more curious about the past history of whistleblower complaints um so we also have testimony from um the f FAA about expertise deficits in inspection and so this is all connected so just to finish up here won't this kill Innovation she's talking about liability she's talking about new steps
with regulatory agencies won't this kill Innovation it's time to Define that term because that's a magic slippery word and when we look at what happens in other Industries things have worked out so the question is whose progress are we advancing is it just novelty for extractive profit that we're building in our Tech ecosystem or is it a case where we're building something that is truly making humans lives better whose story is this whose progress are we centering and you don't need to take my word for the risks here take President Eisenhower he warned us about the emergence of a technology Elite that would present problems for continued governance and stability of our economy and Country
this is the second part of that military industrial complex set of comments that's often quoted but rarely do you see this piece of that speech quoted and I think this is the piece that we need to uh perhaps Ponder in our present moment look at other Industries they have targeted agencies they have enforcement they have ongoing commitments they have various different licensing regimes the level of oversight with many eyes looking at a problem is just completely different from what we have in the tuch ecosystem and the other challenge is that of course as every company becomes a tech company companies that are regulated by these standards will have their own ability to successfully manage their own internal
controls face challenges in light of technologies that are not necessarily in their control in the same way but that they are perhaps naively over trusting with access AI of course will make things even worse you had a wonderful keynote on this yesterday so I'm not going to go through these issues but here are some questions from this so won't these problems solve themselves is the market perfect why don't you trust the market Andrea well um I know a lot of humans and I've been a lawyer for 25 years the Market's not going to fix this and 25 years of security in watching this community and the catastrophe research that I'm doing tells me the market is not going to fix this also
bears bears tell me the the Market's not going to fix this if you have not read this book I cannot Rec Rec commend it highly enough it is the story of a bunch of folks who shared uh aggressively libertarian views who tried to set up a Utopia and things were going relatively fine until bears bears showed up and destroyed everything and so emergent effects and problems matter and arise and individuals won't be in a position to address them and always for better or worse there's always that one guy who let's say finds a dead baby bear cub drives it to New York stages a scene where it was allegedly hit by a bicycle and then runs away or walks away
maybe maybe he walked away drove away I don't know leaves it for someone else to clean up and so in fact that's what happens and that is why the idea of having coordinating points in government is sometimes necessary um and so with that I I will return to my suggested approaches um the law fair post I would love it if you would read and comment to me about what I get wrong I always want to hear that um so in particular what I think we need apart from those interventions is an agency focused on the biggest actors in our economy so um I'm using the heart Scott Rino standard for those of you who are really in the
weeds of competition law so it's the biggest company and uh my ideal situation would be a bureau of Technology safety with three branches an enforcement division a policy coordination technology Futures tracking Division and a pilot projects Division and that is it thank you for your [Applause] attention all right that uh unfortunately I think we're about 2 minutes left before the the break here so uh we won't do questions uh but please uh do seek out Andrea uh after the talk here uh go to her Workshop these sorts if you're interested in these kinds of issues uh the I am the Cavalry track uh deals with these sorts of things extensively and in great detail and and with great results so or
if you have a catastrophe case study you get a sticker all right uh come
[Music]
[Music]
[Music] [Music]
[Music] a [Music]
[Music]
[Music]
[Music]
he [Applause] [Music] [Applause] [Music] [Applause] [Music] [Applause] [Music]
he [Music] he
[Music]
[Music]
[Music]
[Music] track [Music] hey hey hey hey [Applause] [Music]
hey hey hey hey hey [Music]
[Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Music] [Music]
[Music] [Applause] [Music] oh [Music] he [Music]
[Music]
oh
[Music] all right J Chen and Ravid Mazon sponsors I'd like to thank our sponsors especially the diamond sponsors prism cloud and vanta and our gold sponsors Adobe and drop zone AI with it's their support along with sponsors donors and volunteers that make this event possible cell phones we're recording this live on the camera back there so please silence your phones and if you have to ask a question please later come to the microphone so people on the live stream can hear your
question and remember later we expect Applause and good [Music] questions yeah oh the adapters look at the there's adapter there's a white box
nothing okay let me try USBC where's that white box
too many issues we have too many isues yeah
the previous speaker had an issue
use this is no it's not going to work they have like a white adapter like USC in the speaker room there's there there's USBC but there's no yeah there wasn't it was here before
maybe we try m in the end and I give up any there or change the refresh rate on the screen maybe you have the display display mirroring change the display display rate do you know how to do it yeah uh display settings let's display
preset mirror
okay prob we need to use yours it's without internet the will not work you need to fix either the internet or either the
HTM there we [Applause] go and start over
it's on the other screen isn't that mirror uh is this it I know this my homeor no okay uh can we just move it to the other screen where's the is this one is it over here
display settings hang on there's got a there's a mirror
here is this the screen here yeah you should mirror this one
stop this one stop mirroring yeah and and this building
the maybe disconnect it and reconnect it try do yeah do you have the adap there's like a white there was a white adapter for the eight adap I have in the room let's try yes so first of all we don't have Internet and we don't we cannot share the screen as [Music] well USBC speak yeah the white
one can you can you first let me know which Wi-Fi works because no wi it's all the same as ID let me call A please come to okay on your displays or just try what we did there set the display resolution down and then H mirror no but I still don't have a Wi-Fi so this is still a problem so so the SSID bides Las Vegas yeah it's not you're connected but you're not getting internet yeah the staff is giving on the staff Wi-Fi I've got internet problems well problematic it's really I have like a some kind of video which works and it's maybe it working but it's really really slow like I mean it should be pretty fast
they're they're they're actively looking at the speaker Wi-Fi because we have problems in the speaker room specifically with that
AP I guess I can I can give up the the demo but sorry let's see if it works if I'm uh here's a let me try
this there was a white um display port to HDMI box the connection is not working basically and I tested it in the room it tested in the other room and now will go tested with the white display port adapter which we don't have right hang on there we go also the Wi-Fi is not so good which is another problem well that's something I cannot help you with could you slide a this one thank
you this is like a no it's another screen it's not my screen not your screen that's a secondary display secondary you're going to have to change that your settings either that or set it to mirror yep yeah what about um sure this is match yeah uh maybe like wait like this uh I know you can go to settings and make your change or you should be able to go to that's network uh screen display settings yeah display settings I'm going to go to okay and then you can change you want to mirror display yeah use as mirror what do you see change this here is to mirror okay mirror there you
go yeah uh nope what there go okay yeah let's hope everything would be fine see what it's working okay this works one one two okay hey okay in the how's the sound can you hear me good is that how you're going to talk when you present or you going to stand back uh stand back stand back TR from this distance hea one two one two how's it okay I want you to lean in a little bit okay be aware can I hold the mic can can we just hold it walk hey we have a wireless as well would you rather Wireless and not this I'm good with it this you're okay with this so we will get you a wireless stand
by check one two
I'm okay with this uh but you can't walk around what here do this that way you can walk around okay put that in your
pocket you don't mind
and then this is on off so that's on hey hey yeah can you hear me good okay okay can we start yeah
I thought you were it's it's worth it I'm telling you it's worth it wait for
it thank you okay talk try it okay yeah can you hear me go okay thank you okay great great so let's start um so thank you all for joining wait wait wait
um what's now check check check check check can you 1 one 2 three 1 2 3 4 yeah one to three check keep going check check check good good good can you hear me okay 1 to three are you ready to switch over live we are live okay thank you very much thank you thank you okay let's go um so thanks you all for joining uh we will introduce today bolab Buster which is our methodology of um automating Bola detection vul abilities uh with using llms uh but first thing first let's introduce ourself so my name is Ravid I'm a senior security researcher at Pao Alto networks um I'm part of the was team which is web application security and
API security and in my free time I like watching football games traveling the world and take care of my dog uh Maple sorry for delay apparently AI doesn't solve everything we still need human in the loop a lot of human thanks for the technical team and my name is Jay I am a security researcher with pal networks my research has been focusing on identifying the risk and threat in Cloud environment recently I have moved I have switched my research Focus toward generative AI in particular studying the potential malicious use of generative AI when I'm not working I spend most of time with my hyperactive twin boys who behave just like minion and I also have
two cats when my kids go to bed I become cat slave feeding them cuddling with them and cleaning their little box okay um so let's quickly go over the agenda for today so we will introduce the concept of Bola we will see our methodology of automating Bola detection with llms we will see an actual test um that detects a real Bola and eventually we uh we will show you how we hunted down 17 new ball of vulnerabilities in the while and what lessons did we learn uh during the process um so Bola or broken object level authorization uh um first of all our motivation for this re for This research um I'm not sure if you are
familiar with Bola but it is the top risk at the the O fpi top 10 it's the number one and it's also the fourth most reported vulnerability in Aker one so it's very popular it's very severe and there is no automation tool that actually detects uh Bola in scale today um for all of these reasons we decided that we need to develop our own methodology uh to be able to automate the B detection and uh solve and unsolved uh question if you take a look uh in this screenshot you can see a patient uh application um which a patient can query an API call uh and use his visit ID and it will get his doctor uh
noes but what would happen if the same patient will try to query another visit ID that belonged to another user um if you will be able to fetch this the sensitive data of another user we have a authorization issue or Bola so Bola is a basically vulnerability that arise when the application failed to validate the if the user is authorized to access modify or delete object that does not belong to him so imagine that I'm able to delete Jay's comment from Instagram or Twitter which obviously I shouldn't be able to uh this is an authorization issue and we will have a ball of vulnerability and the consequences can be data leaks data manipulation or even
full account takeover these are the challenges that we faced during the process of the development first of of all uh today's application has multiple users typically multiple roles and resources and it's really difficult to to understand which user is allowed to do what secondly most application today are stateful which means that every uh action you make every endpoint you call affects H and change the the state of the application so imagine that you try to delete a comment from an article first you will need to create a comment sorry to create an article then to create a comment and only then you will be able to delete it so there are some dependencies between endpoint that it's
really hard at first to recognize also um there is a problem of lack of vulnerability indicators so imagine xss SQL injection they all have a pattern we all know how to recognize them pretty quickly Bola is a logical error um it's really it doesn't have a clear pattern so it's really um difficult to understand whether it is a b or not and lastly uh the context of the application it's it was difficult to understand exactly which endpoint or parameters return sensitive data and what is the actual impact of each action that we make so all of this um this was a real challenge to automate the the detection of baa in in scale um and
yeah thanks RI for covering the background I hope everyone Now understand why automating Pola detection is not easy so Pola is not a new problem it has been around for so long ever since we had internet however it was only two years ago that we realized that AI might give this problem a of Hope in particular the rapid advancement of AI give provide us with new tool to solve problems that were not possible previously in particular it also happened that the challenge our challenge of extracting context and logic information from texal data is what large language models are extremely good at so that's the beginning of our journey in using AI to solve this problem here's an high level overview of
our methodology the only required input is open API an open API spec or a swager spec in the first stage we identify the end point that can be vulnerable to Bola not every endpoint can be vulnerable to Bola we use AI to help analyze every endpoint and its parameters to select a subset of end points this step help us focus on a smaller relevant endpoints only and avoid wasting time on the end point that are not at risk so it's important to know that we call this we call our Target endpoint as potentially vulnerable endpoint short for PVE I will switch between Target end point or PVE or uh potentially vulnerable endpoint in this in this talk
the next stage uncover the dependency relationship between each endpoint modern web application are complex with one endpoint depending on many others for example if we if I want to test uh an endpoint that update an invoice I first need to call the endpoint that create invoice and before I can call the endpoint that create the endpoint I also first need to get call the endpoint to create some transactions that can be included in the invoice so it is crucial to identify the dependent end point of a Target endpoint before we can accurately test it with the endpoint dependency relationship identified we can then calculate the execution path to each Target endpoint to each potentially vulnerable endpoint we then create a
test plan for each Target end point the next stage then turn the test plan for each PVE for each Target endpoint into a set of executable bat script using large language models and there may be one or multiple t uh execution path to each Target endpoint we aim to cover as many path as possible in the last stage we set up an actual API server and run all the executable Bash script to send the actual API request to the Target endpoint the process of user registration user login and token refresh have all been automated and we also use AI to help analyze the logs and response during the test to determine if a end if an endpoint is
vulnerable to Bola now let's dive into more detail in each stage the first stage identify the end point with input parameter that reference private sensitive or confidential information these are the end point that we primarily focus on these are the target endpoint potentially vulnerable end point let's use the first endpoint as an example the parameter username here indicate that this endpoint May reference to some data associated with one with a particular user as a result if this endpoint is vulnerable to Bola and the T haer could reset another user's password similarly the second end point here if it is vulnerable to Bola and attacker may be able to change another user's input uh email traditionally pentester manually
look through every end point and its parameters to identify identify their target to identify their target end point this process has been slow and combersome especially for large application with hundreds of end points we leverage AI capabilities of reasoning and understanding task to automate this step here is the snipet of the prompt that we use to communicate with AI basically this Trum this prompt instruct AI a set of rules and example to identify parameters with uh that may reference to sensitive information the AI then return us the end point and parameter that meet any of the conditions here the next stage uncover the dependency relationship between endpoints as I mentioned mod modern web applications are complex and stateful
meaning that the EXE the execution of any end point can change the state of the entire application and affect the outcomes of other end points that's why it is crucial to identify the dependency relationship before we can accurately correctly test any Target endpoints in this diagram the endpoints on the right are we call them consumer endpoints and the endpoint on the left are producer endpoints this is one of the most important Concept in our research to in order to identif y the dependency relationship producer endpoint on the left output the values that consumer endpoint need as input again the producer endpoint on the left produce output the value that the consumer endpoint need as input let's use uh this
as an example the consumer here is delete username and in order to correctly test this endpoint we need to fit the end point an existing and correct username if we fit in a random username the test case will always fail and give us meaningless results in this case this consumer a coin has four producers and each of these producer can out output the existing valid username that the consumer endpoint can use for testing each Endo can be a producer a consumer or both let's look at another example here the consumer endpoint here is delete comment it has two required input Slug and common ID and they can all conf from its producers get comment and CST comment in turn these two
producers also have the dependent the required input Slug and they have come from their producers and here's the snippit of the prompt that we use to teach AI to recognize dependence dependency relationship between end points it is important to to to to know that although uh this process can be done to htic asking uh using heris to match the output parameter with of a producer with the input parameter of a consumer this heris matching algorithm is not reliable for several reason first developers may use different parameter Nam to reference the same data object within the system and developers may also use the same parent parameter name to reference different object within the system that's why we need to use AI to study to
analyze the description in the spec and match the end points by their functionality rather than just the par par par
name the next stage turns the pairwise dependency relationship we identify in the previous stat into a dependency tree for each Target end point in this in this dependency tree the r note represent our Target end point tve so we create one uh dependency tree for each Target end point and within this tree any two directly connected end points represent a consumer and producer relationship in this diagram the PVE is the parano and it is the consumer and .1 is the uh child note of PVE and it is producer again producer out put the value that consumer need as input and let's use this one as an example .1 is pv's producer and n.3 and four are n1's
producers in the next stage with with all the dependency re relationship figure out in the next stage we calculate the execution path to each Target end point in the dependency tree a path from a LIF node to the root node represent a depend uh an execution path and there may there can be one or multiple execution path to each Target end point let's plug in some real end point to show how we calculate the execution path in this case the PVE is delete comment and it has two producers get common and post common and this two producers in turn also each has two producer so in this case our Target end point DD common has two has four
execution path finally in the last stage we turn each execution path into an executable B script and run the B script to actually send API request to the Target end point the process is more complicated than just generating and run in the script as RIT we expand in the next few slides thank you Jay um so before we going to see a real balla test and a demo let's see the the rest of the stages before we actually can generate test so first we will need to register the users and collect the login data these steps also is done by AI with a human verification first we will generate uh the users credentials that meets the the
criteria AI will uh analyze the open API spec and and understand what are the complexity of the username and password that uh a user can have in the application next we will create an execute the users registration phase and lastly we will fetch the login data and Save in a dedicated uh file for further usage and before we can actually create the test we need to isolate the test data open API spec can be really really huge and we don't want to deliver to H to feed AI with a huge API spec we want to cover only the specific data that is relevant for each test so Jay mentioned we have a consumer and producers for
each test we will isolate only the data for this consumer and producers and create a new trimmed spec file which we will feed AI we do we do it for efficiency we want to have only accurate data to prevent AI from having a mistakes and for cost we want to have H the least amount of token that we can send AI so when we try to generate a test script it is saved as an executable bash script you can see an example of a really simple uh ball test we have a consumer of put a user password and the producer will be get users which will get the username this test will be saved in a
put get directory put will be the parent get will be the child and the test generation script actually runs asynchronously in order to save time and right now we have an average time of 1 uh 5 Seconds to generate each execution uh B screen lastly uh we do we perform a when when we run the test um on the application we do it in a certain execution order and the main goal is to avoid avoid test failure due to technical reasons we want a test to be failed only if the PVE is not vulnerable to Bola which should be so we first run the test that populates data and resources to the application and only then try to fetch
them to eliminate the to eliminate the fact that uh to fetch a nonexisting uh resources and we try to push the delete and update uh operations to the end we don't want to delete a user and then then try to to fetch it because the test will fail due to a technical issue and after we finished creating our methodology we uh actually made the evaluation process so we evaluated babster against westler wrestler is a open source API fuzer created by Microsoft um it's one of the best API fers that you can find at least open source so we wanted to test our methodology against the best and its goal is also to automate the testing of services through rest
API um and find security vulnerabilities basically and what we did is you can see this table we took three application this is open source application uh vampy capital and crappy all of them are uh deliberately uh vulnerable to the OAS SPI top 10 so they have existing Bas um you can see the number of Bas in in each one of them so again we tested restler um against our tool and this is the results of uh the wrestler run wrestler couldn't discover any B of vulnerabilities in in either one of the application um I will just say that we use the default configuration and they claimed that they do uh find Bola with the default configuration but they
didn't and the number of API calls you can see was pretty big thousands and even uh hundred of thousands calls which making a lot of load on the application obviously and this is our results which were amazing we found all of the Bas in all of the applications and we did it with less than 1% of the amount of API calls in comparison to restler so this is really huge um in terms of the loow we didn't even make any load on the application and we were able to find all of the boras also we focused in on the true positive rate so our goal was to uh if an application have aola we will want to
find it um we we didn't really care at this stage about false positive because it's really basically impossible to avoid them but again we had the 100% true positive rate so we found all the Bas in the application and now let's see an actual Bola test so this is an high level example of a test uh we have here two producers and one consumer which is the PVE so in this scenario Alice will try to create an article she will create a comment for this article and eventually Bob will try to delete Alice comment which is the potential Ebola so first of all um Alice um and Bob will log into the system and they
will get a unique uh token afterwards the sequence of the test will begin uh Alice will create a new article in the system we will save the article title later tic later Alice will create a comment for this article we will save the comment ID and lastly Bob will attempt to delete all his comments and this has two options uh of results so if we will be able to do it we will get a 200 okay and this is a potential Eola vulnerability for this end point um and if everything is correct and we have defenses it will be of course forbidden let's see how the code looks like so we have here Alys uh which
create an article I hope you can all see it uh first of all we will create a unique random string so we will try to use it as the article title to have a unique title every time and then we will create create a post request to/ articles in in this uh in this case to create a new article you can see that the authorization Adder is user a token this is the ident identify that Alice is making the request and we will use the random string as the article title so this is the the API call we will save uh the article title as a slag and then Alice will try to create a
comment for a article and as you can see here the the if you can see the slug which is dynamically being used in this API call um this is the slag that we saved from the the previous uh API call so we are using the article she already created she create a comment and we save the comment ID and lastly Bob will try to delete uh air comment so you can see the API call is being we are using slag and comment ID the one that we saved before now you can see that the authorization Adder is user B so this is the identifier of Bob instead of Alice and basically in the end we will
check if the test passed so just one sentence about the the the check in in this case in our case it's enough to to Mark a test as as a Bola a potentially B at least if you if the PVE rets 200 okay if you think about it every test we make is is malicious by definition we the last request is a user trying to perform an action that it should shouldn't be allowed to so we expect not to get 200 okay if we do we can mark it as a high potential for Bora and later on WE perform a human analytics and checks to verify if it is a Bora okay so for the fun part let's see
the cves that we found uh and before that okay I'm I'm not sure the demo will work no it's not working I'm sorry we have a technical issues but let's see the Bola that we actually found um in the open source application so first we have Harbor which is a cloud native container registry um it's basically equivalent to Docker house I'm sure that all of you know dockerhub um and it's a cncf graduated project so it's very popular it's being used by it's being downloaded by 2 million uh 2 million times and we found uh in 2024 we found e Bola there we have also grafana that I'm sure that most of you are familiar with it's
a very very popular data visualization and monitoring tool uh it has about 20 20 million users around the world and we were also able to find the baa there and lastly we have easy appointments which is a appointment scheduling application um it is less popular but it has almost 200,000 download um and we will managed to find 15 new Bas there um seven of these vulnerabilities are targed as critical so they have a CVSs score of 9.9 which is the highest so imagine that seven vulnerabilities out of the 15 allow uh a full uh compromise of the application so you can do whatever you can be an admin basically you can do whatever you want uh so this is this was a pretty big
uh a big achievement for us and let's Deep dive to talk about the Arbor vulnerabil so Harbor actually has a projects and the feature that we found vulnerable is the project configuration metadata um every project have users so you can be an admin you can be a maintainer a developer a guest or a limited guest and I hope you can see in the screenshot that uh this is the configuration of a project in Harbor um you can do a bunch of stuff you can change it the project to be private public you can create um Autos scan for image vulnerabilities and much more so Arbor claimed that only the project admin can create and modify or delete this
configuration which make a lot of sense this is crucial part of the of of a project but we actually found out that when we are logged in as a maintainer uh first we try to modify these attributes via the UI we we were not able to do it as should but we found that we can do it via API so there was a discrepancy between the UI and the API which allow us to to an unauthorized uh user basically to create edit or delete the project configurations so the issue here is the maintainer actually extend his Privileges and now we can make a private project public deploy unverified images and bypass vulnerability scanning and and
more um the consequences can be really bad so we you you as a malicious malicious maintainer you can compromise the entire project Integrity um and the security posture basically so Harbor recognized this vulnerability and issue a cve uh it was not long ago basically like two weeks ago and and um they publish the details in their uh security advisory so if you want to take a look it's open source so you can go and do it and also if you want to read more about the technical details about this vulnerability and our uh uh methodology uh go ahead and scan this code this leads you to our blog that describe this vulnerability yeah thanks AR I hope everyone
understand by now why we need Ai and how we use AI we started the project by dropping an entire open API spec into chat gbt and ask it to find all the Bola end points as you may imagine the result were not great and we not only exceed the token limit but also confuse AI a lot and so there after many many trial and errors we gradually learn how to to collaborate with AI to optimize its performance and these are a few most important lessons that we have learned throughout the research first AI isn't always the best solution AI should not be used to solve simple problem with humanistic solution like sorting finding path solving equations this problems
have existing efficient and Optimal Solutions although we may use AI to solve these problems but usually it does so in a much at a much higher cost and longer time remember don't shoot a mosquito with a Shotgun If and if a problem with existing heuristic solution always choose heuristic over AI second don't trust always validate blindly trusting the output of AI can be very very dangerous especially in application in critical application where mistakes can cost millions or even human life in our research we often generative AI often give us non-existence parameters end points or sh command in our case these mistakes are not life threatening but they result in failed test case false positive and false
negative as a result it's so crucial it is so crucial to always validate the output of AI before using them or passing them to the next stage lastly makes ai's job easier treat AI like a very capable Junior colleague who can do simple task extremely well but it can start making mistake if the TX get more complex it is thus the human supervisor responsibility to simplify AI task in our case it was a bad idea to just give the entire API spec of thousand of Lights into AI it confused AI a lot as a result we break each API spec into many many smaller pieces and only fit AI the relevant piece the relevant pieces of
the current task when working with AI divide and conquer is always is always a good strategy although we have seen some promising and successful successful result of This research we still have some remaining Challenge and room for improvement first our methodology is very sensitive to the quality of the input API spec currently we treat the API spec like the absolute truth and build our entire test plan based on the API spec however throughout our research we found that many open API specs many API spec of the open source project are outdated or inconsistent with the actual API functionalities this inconsistency result in a lot of issue fail test fail execution path and and false posi false
positive and false negatives next not every API application out there has an uh API spec available some appli some application don't have maintenance or even documents and there are application deployed in more restrict environments such as industrial control system in which we don't have direct access as a result in the next phase of the research we want to explore more data sources such as uh pcap flow log or even source code to help AI understand the application lastly using generative AI model can be quite expensive especially for the more advanced models the me the cost of our methodology is proportional to the size of the API spec and the complexity complexity of the application luckily
the rapid advancement of AI and the intense Market competition between AI service providers the cost of AI has reduced a lot in the speed that faster than we could imagine compared to just 6 month ago the cost of testing and application has been cut by half while the performance and the speed of the model we use have all significantly improved one amazing side effect of working working with AI is that our the performance of our application always get a free boost whenever there's a new generation of model available and let's conclude our talk I don't know if you have time for for a question but we can talk after [Applause]
after thank you very much thank you thank you [Music]
[Music] w a [Music] [Applause] [Music] [Applause] [Music] I'm [Music] just I'm just try to give you [Music] something I'm just tring to give you something I do you I'm just TR to give you something [Music] he [Music] [Applause]
[Music]
[Music] [Music] I'm just trying to get I do I'm just tring something [Music] I'm just trying to something okay [Music] I I'm just trying to give you something [Music] w
[Music]
[Music]
[Music] [Music]
[Music]
[Music]
[Music] [Applause]
oh [Music]
[Music]
[Applause] [Music]
[Music]
[Music]
Vegas we'd like to thank our sponsors especially the diamond sponsors pisac cloud and vanta and our gold sponsors project circuit breaker and semrep it's their support along with sponsors volunteers that make this event possible cell phones remember to silence your cell phones we're being streamed live on YouTube so remember if you want to ask questions later please use the microphone so the people on the live stream can hear your question today's talk is jit happens and we have Matthew Sullivan and Dominic zenari thank you so much awesome sorry we're having a small technical issue still one moment as we work through it all good good morning everybody I think it's still morning um want to introduce myself my name is
Dominic zardi um I'm on the sharing the stage with uh my friend and colleague Matthew Sullivan who uh I join on the infrastr infrastructure security team at instacart um so I want to thank uh want to thank the audience for spending some time with us but I also want to thank uh bides for uh giving us the opportunity to sell you a yacht not an 80ft cat ofar like we'd all love uh but this is yet another AI talk um we're here to uh most most at most security conferences uh AI is viewed in a negative light we want to provide a positive light we're not saying llms are a magic wand but they're used to greatly
enhance uh security tooling uh this isn't a paradigm shift but it gives small teams the breathing room we' we need uh because the Cavalry isn't coming but the controls are we have a lot to do security team workloads are growing um Am I Wrong is anyone bored at work today um we need to be faster and more agile every day we hear do more with less there are always more controls who's going to write all this
junk Ron he knows what's up so we have a couple options uh with all of these incoming requirements uh we can level up be the best team we can be be the best in the world or we can turn to our robot overlords um and reach out to Skynet because we also believe in work life balance so automating the gray area um when I think about security automations uh I think about binary decisions most of the time when we're setting up automations we deal with true false we deal with static data sets llms can help us in the gray area where context is everything um where humans might have to spend hours pouring over logs pouring
over giant data sets uh just to make decisions let's say your audit team asks you we need users to only request appropriate access every single time before AI this might be a look you give your auditor I don't know but with AI let's vacuum up the previous audit logs let's vacuum up the history that's been provided um and recommend the appropriate role to a user based on what they've actually used another one we need you to review and update your role description when there's a change in privilege audit will ask us to regularly update our role descriptions whenever something changes before ai go through every I am Ro look at the policy statements update the description destion for them put
them in a spreadsheet provide that as evidence with AI let's instead take those steps pull the policy statements to a role send those to an llm and have the llm write the description for you it's not difficult it's just tedious why not have ai do it so uh before we demo this uh just quick crash course and in prompting an llm uh they have two query components uh the template or system prompt is the first one uh we want to make sure we provide accurate instructions to the llm uh so that it knows how to operate the user prompt following it typically provides the content uh that it needs to follow those instructions um quick reminder um it's a
good uh thing to not that when you're comingling instructions and data you want to make sure that um you lock it down if the data you feed to the llm is provided by the user so R descriptions from AWS apis um let's look at how we'd use a tool an AI driven script uh to generate Ro descriptions from AWS APS um you're going to see a side by-side display of data on the left are the prompts that we send to the llm on the right is the response from the llm so let's generate a roll description we'll switch that perfect so we're going to kick off this python script as I mentioned we're using a
library that splits data down on the left is what we are sending the llm I am an internal auditor provide your persona right away this helps build the output you get back if you want to appease Auditors say you were an auditor um we give more information we're basically saying please provide a three sentence summary of what can be done with this role so we can use it stand by for list of actions in this case this is a live PLL from a demo AWS environment we're literally pulling the um IM am policy statements from the role that we've sent and with the system prompt and the user prompt together we're going to get a response
from the llm based on those policy statements saying this role allows users to manage and interact with the S3 sqs you want it to be um you want it to be descriptive enough to approve be approved for uh Auditors on the outside from an Evidence perspective you want requesters to understand what they're asking for without showing them a policy statement um and you want the approvers to know what they're authorizing So This is highly visible in three spaces um at instacart we use a platform for Access requests known as conductor one um that platform is completely based or at least in our configuration we base all of our config in terraform and identity as code so this piece can
easily plug into our terraform files so the next step here that one is to take this information and literally pass it into a poll request so not only are we generating it we're now sending it to GitHub my windows disappeared you're good Comm yeah perfect so as we open this up as I mentioned all of our configuration for conductor one the access request platform is based on uh terraform files this Ro description used to be an example Ro that does a little this a little that we now have upto-date information on exactly what this is you can change the configuration to be as short or as long as you want um but right here we can review the changes
approve it and move on all audiences uh are happy Auditors requesters and the authorizers so so's here to walk us through how instacart's uh using these Concepts in production to solve more real world challenges solving the pains thank you thank you thank you I'll have you hold my speaker notes I'll trade so that's a really simple example right we're pulling in data we're throwing it to an llm it gives us a description you can probably start thinking about some really interesting use cases you might have though are there times in your business when there is some complicated alert I know the cloud um tooling that we utilize the alerting is typically very hard to understand and if that triggers some
sort of event that comes to your phone what's helpful the Json blob or hey it looks like a new IM IM user is doing some weird things in your AWS account right even using these simple Concepts it can actually really help be a quality of life change and so these are the types of things that we're working on at instacart is even taking our alerting running it in through for us we use tines a tines workflow that just makes it easier to digest the other thing too is then my on call rotation just doesn't have to be so scary to a new hire you know I remember when I started on call it was a lot to take in and if it's more
simple English then it allows somebody to feel a little bit more relief as as these pieces of data come in and out so again it's really important to us that you look at these things not as some scary new technologies that are going to ruin our lives but like well what if we did embrace them and what if we actually made it so that these were a quality of life Improvement for us so we didn't feel so burned out at this point I'm going to talk about how we solved this real world problem we wanted to go public uh so I started instacart I accepted my offer on the day we filed our S1 which is our intent to go public
with the SEC uh what I did not realize is that basically from that exact moment until we went public in September of 2023 that is all I would do was just that work uh and in case you're wondering no they didn't tell me that is what I would be working on but that's it's okay I really enjoyed it um so a significant amount of time was really spent trying to figure out how to build an access control program that could meet our objectives and we could be proud of you see when a company goes public in the United States your stock offering becomes subject to sarban Oxley of 2020 or 2002 uh and what we
could do is bore you to death or we could not do that and simply say that basically this means that you don't cook the books and I as instacart shareholder have come to the conclusion that's generally a good idea not cooking books so socks dictates that we will have the right people uh that have access to financial data only authorized persons this makes sense we don't want people one we don't want people seeing materially impactful financial information or compromising the Integrity of that financial information makes sense you see there was this company called Enron Enron lied about their financials investors and even some of their own employees lost everything jail time you get the picture the
entirety of socks can be distilled down to saying boy that sucked let's not do that anymore Soro so socks takeway access matters it's the only thing I really want you to take away about socks doing socks means that you generally care about access all right so we've got my mandate and my deadline IPO is coming uh what can I do to make this Access program great because I have this thing I don't know if you guys are like this where I don't like working and I would rather do literally anything else all the time but because I have to work in order to survive then I do it too hard right and so if I'm going to go in and
make an access program it's going to be awesome right uh so one of the things that we settled on early is okay we'll use Justin Time access jit um and basically in jit you grant temporary permissions as Dom mentioned a second ago we did a evaluation and ended up purchasing a commercial solution in this space uh who we felt was aligned with our values of kind of like disrupting the status quo a little bit um so we fast forward and uh now we've got our roll out of that done we've got one roll onto this J in time flow it's kind of like an administrative role that our infrastructure team uses um but we hit a
snag though as we quickly realize that manager approval sucks that's really hard you see the problem is your manager is busy they lack context they lack expertise sorry managers you do some managers just turn 50 and are out of the office trying to find themselves by backpacking across Europe and most people just buy a sports car and get a dog but what do I know so we need to do some real talk about how we're going to fix this spice level 100 here we go let's talk about doing everything everywhere all at once with access requests only instead of that it's everywhere all at once is on fire everything is on fire and everything is
terrible in the status quo in our industry and I think it's time we acknowledge it security and audit continue to double down on the existing way of doing access requests whether that's meeting socks or if you're a Fed ramp shop whatever it is you're doing the same thing we've been doing for the last 15 years possibly maybe you're maybe you've improved and that's great the problem is the model is fundamentally broken we're made to do it anyway though we have to do this thing we submit the ticket and we wait a week for a manager appr wait a week for the owner to actually action on the request and then you defend your access until
they pry it from your cold dead hands because it took so darn long to get it in the first place you will never let it lapse somebody walks up do you still use this oh every day yeah I keep prod alive uh-huh being able to hit that S3 bucket is keeping prod alive I gotcha right the fact of the matter is your access requests look like this graph when it comes to manager approval denied 0.1% approved 99% .9 and it's because they click the wrong button tell me I'm wrong our industry has a bad case of the not my problems when it comes to access approval with ask ask your manager for approval and ask system owner for
approval we've solved access requirements as a checkbox compliance exercise and nothing more we haven't added security so I thought to myself since I'm spicy let's not do that how are we going to fix it and I figured it out we simply had to kill all the humans we could do that though by pre-approving access our identity governance tool allows us to provision temporary access right so what we did is we set a maximum time of 90 days or less for really important things and we allow users to renew that immediately as long as their user is part of a pre-approved set of attributes that we keep in the terraform that we were just talking about so Dom actually spe headed
building a uh a terraform provider and we want to share that with the world it is specific to conductor one the tool that we use at instacart um but this actually is uh a terone provider that kind of puts conductor one into the instacart way and so we've built a kind of an abstraction layer it's opinionated um but we think it works well and most importantly maybe it can help somebody else and if so fantastic so have a look at that if uh you happen to be a conductor one shop or you just want to see how we're thinking about this problem uh for whatever tool you might be using so we've done it everybody we
solved access hooray okay it's not actually that easy we solved a good portion of it but the problem is that some things aren't cut and dry this is not cut and dry we have developer power user developer power user can basically write to uh high impact S3 resources and it's not clear how we would build rules for that so we had to set it up for manager approval uh you see the condition for needing this role as far as we could tell in our role engineering was have you worked here a long time are you important do you get involved when things break that's a hard set of criteria to add to your workday profile how are we going to solve that
we need another robot so we need to build a robot we wanted to develop a process where we could trust uh the automation of justtin time approvals instantly in Risk appropriate situations we want to leverage an llm to do what it does best to take in a huge amount of data and help us sift through it I must be clear I know you're already thinking it no we do not ask an llm should we approve this don't do that that will go poorly for you what we can ask the llm is look at these people who already have access to this entitlement look at the person who's requesting access are there similarities between those two
things and llm is really good at that taking a huge amount of data and sifting through that I can do that as a human but for an entitlement that might have a 100 200 people in it that's going to take me a long time with an LM it takes seconds we also optimizing the enduser experience we can get rid of a bunch of significant complexity from their side of things it just looks like they put in a request and it gets approved and they get access they don't realize that this entire crazy thing has happened behind the scenes to make that occur so so let's talk about automated access approvals we built Gadget and in case
you're wondering yes we open sourced it we'll talk about that in a second um Gadget is our answer to this problem and most importantly it's our answer in a generic way um what we didn't want to do was solve this just for ourselves and then just kind of vanish Into The Ether so we've built a pluggable interface for interfacing identity information llms and IG tools um we're pretty excited about it and we'd like to show you a demo now about an hour ago I submitted an access request uh and everything that you're about to see is actually real in production which means that it will fail and the demo will suck when that happens I have a recording um so my access
request is for the AWS role that I use for my daily job uh I'm going to go ahead and run this utility it's going to take a look at that access request and what you're going to see is that left and right Viewpoint again of interfacing with the llm so we're going to see how we prompt the llm uh with this gadget utility because it's going fairly quickly I'll just scroll up system prompt here is that we have a list of employee IDs uh and we're creating examples for the llm so this is a very verbose example I'm sorry it takes a lot to prompt an llm to do these things but we basically say a new applicant would
want to join your group and for example if they were an analyst in online grocery which of these two job titles is relevant and it would answer well the staff engineer and online grocery is the most relevant versus a compliance auditor if there's no overlapping information just return nothing we take that system prompt and send it and now we send our user prompt so this is my real data I am a staff security engineer whoa boy I'm a staff security engineer on the in hyen security team take a look at all of these titles of people who also already have this role and compare it and we get a fantastic match right away that match happens to be Matthew
laurore senior or St staff engineer on my team so we're doing good we've got an exact match great next thing we're going to do is Rerun that same logic and we're going to ask about organizational units many companies have different organizational units that do different things let's take a look at that relationship between the OU I'm a part of and the role that we're requesting same thing we prompt the llm and then we start feeding it real world data out of that is going to come a lot of really strong matches effectively this is going to say hey there are a lot of people in that role that match your organizational unit very well what are
we doing with this data each time I get an answer uh literally in the code base it's going into uh a score a numerical score that's going into an array and just being stored we're going to run a computation function in a little bit so we're not making decisions yet we're just getting some results uh in the real world this is actually something that we fire off all three questions at the same time and a multi-threaded pool and then get all the answers back at once and check for the result the final thing we're going to do is to actually take a look at my uh entitlements name and description a second ago Dom just told you that we can
do this crazy thing where we look at all of the entitlements access we can generate a description a human reviews that make sure that it seems correct we load it into our terraform we pull it right back out in this tool and then compare my job function to the description on the entitlement the AI will take a look and see that I in security am requesting the security role and it's a role used by the security team what's the relationship score 1.5 this is the one place where you could be like what if the AI hallucinates because I've asked it to generate me a score so we have to be careful with this data this could be bad data I will admit so
how did we check that first we ran this through three months of data back and replayed it all from real access requests and we manually looked at every single score and I said does that look logical and it was within the realm of being good enough in 100% of cases and it was perfectly accurate and about 95% of cases so we just determined that was plenty good enough for our cases the other thing I need to remind you of is our goal is to just not approve everything if we've done that we're already better than the status quo right so we can have a little tolerance for you know failure here all right uh so
now we've taken those three scores we add them together I get a 2.0 anything above one means this is probably a good match Gadget has a configuration option which allows you to say which entitlements uh allow automatic approval and addition to those roles if you're going to do the manager thing because maybe you have a high impact role and you don't want to do auto approvals Gadget still comments and that can really help your managers so of course you have that rubber stamp problem right but we add a comment that says we really recommend you don't do this take a close look understand what you're being requested understand what you're approving understand whether or not the
person who's requested it really needs this as part of their job function and we found an incredible amount of success with that so even if we don't do auto approval just the guidance piece we've seen immediate value at instacar which has been pretty awesome head back where is my display I've lost my slide deck just reopen it all right again one of the things that was really important to us as we started this journey is I hate when companies take more than they give from the community and this was something that I thought we could really try to change the way we do things in our industry and so we're really excited to be able to
open source this I beg you we would love to hear from you I want to see PRS I want to see you add more tools um and and this is a very uh like I said very pluggable interface if you want to run a different llm we use open AI you want to use anthropic that's supported write a plugin for if you want to use your own IG tool that's a plugin support it if you want to write your own scoring mechanism and not rely on ours fantastic you can write it at as a plugin this is also able to support running as a Damon or as a Lambda or as a web server so if
you have a web hooking capability in your existing IG tool it's supports that out of the box as well at this point we're going to go ahead and stop and ask what questions the audience has we really appreciate your time thank you so much for [Applause] coming and I'm expecting good questions because this is kind of a spicy topic y microphone's in the middle seems like the mic is off
now the mic is on perfect that's good that's a great use for llm thanks for the talk in the in the tool especially open source Tool uh I'm imagining you didn't start with zero access you started with a company that has a bunch of stuff that you can use to go tell whether the current request matches past requests um have you thought about how to do this in like a bootstrap situation right like we're going into a new cloud or first time we've interfaced with this tool or those sort of situations you know I'll be honest uh I did not think about how you would start from zero to one just because I was starting at like
99 and going to 100 but uh I think that there is a really interesting opportunity to do some incredible things with let's say you're really doing zero to one no company's actually at zero right so we can come back to what Dom was talking about if you can write some basic things that say take in some cloud trail log see what people are actually doing and then maybe that can help you with your role engineering ing that's great I mean there are companies out there trust me I know we're customers of them that want to charge you $200,000 to tell you what your people are using and then provide you a good role template to
use we do that in 40 lines of python right so there's an opportunity there to just be Scrappy and um and I think for me that returns me to my roots I don't know about everybody in this room but like I started as the one security engineer at a startup and it was horrible and I had a ton of fun and did it for 10 years right and like I remember those days we didn't have budget for products you just build some shell scripts and prey and it was enjoyable to kind of get to look at this situation kind of like that be Scrappy be efficient and fast and how much progress can we make with as little time
as possible and I think we achiev that goal anything you want to add yeah I would just say just to Echo that when you have cloud trail logs feed them in and tell it hey generate me a new role we've done that at a in a couple of instances so it works out yeah can't stress up human in the loop is still needed I mean you need to be sanity checking these things it does not always spit out valid terraform you so human in the loop but it's it's again it's a tool yes man I've never seen before from instacart um hi uh what do you do if you have Auditors who even though they're
not supposed to ask for absolute Assurance they are asking for absolute assurance and they're going to ask you and I know they're going to ask you this because I work with you uh how do you know for sure that it's not hallucinating like what additional checks can you add Beyond reverting back to a horrible manual user access review well assume that you're going to do a garbage like you know normally you'd say garbage in garbage out we're going to start garbage out and then check garbage in right so take everything all those IDs that you just saw spit out back to us double check those go back to your same list of users make sure that the
data that you fed in stayed the same the IDS are still the same the job titles you fed in came back out an llm isn't going to start swapping those it's going to hallucinate them and so you have good signal right there to know if what it's generating is real or not real um and then the other thing too is you have to ask just super basic questions you saw the structure of the Json that we request back it is dead simple simple structur simple key names we're not doing like nested Json you know responses you can't the llm will freak simple questions it's like talking to a four-year-old I would know I have one
so sounds
good thanks again appreciate your time today [Applause] [Music]
oh
[Music] h
[Music]
[Music] oh [Music] [Applause] [Music] [Applause] [Music] oh [Music] [Applause] [Music] I'm just I'm just dring [Music] something I'm just tring something I do I'm just TR to give you something [Music] m [Music] a [Music] [Applause]
[Music]
[Music] [Music] just TR to give you something I'm just TR to give you [Music] something I'm just try to I do I'm just trying to give you something [Music] a [Music] he [Music]
[Music]
[Music]
[Music] [Music]
[Music] is
[Music]
[Music] he [Applause]
right
[Music]
[Music] [Music]
[Applause]
[Music]
[Music]
n [Music] a [Music] l [Music]
[Music]
[Music] n [Music] [Music]
[Music] [Applause] [Music]
[Music]
[Music] n [Music]
[Music] [Music]
[Music] [Applause] [Music]
[Music]
[Music]
[Music]
[Music] [Applause] [Music] he he he [Music] [Applause] [Music]
[Music]
he he
[Music]
[Music]
[Music] aack
[Music] hey hey hey [Applause] [Music]
hey hey hey hey hey hey [Music]
[Music]
[Music] [Applause] [Music] he [Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music]
he [Music]
[Music]
[Music] [Applause] [Music] he [Music]
[Music]
he
[Music]
[Music] h [Music]
[Music] [Applause] w [Music] [Applause] [Music] just to something I'm just TR to give [Music] something I'm just I do I'm just trying to give you something [Music] w
[Music]
[Music] [Music] I'm just to I'm just trying to give you [Music] something I'm just trying to give something I do you I'm just trying to give you something [Music] m [Music]
[Music]
[Music]
oh [Music] [Music]
[Music] he
[Music]
[Music] [Applause]
oh [Music]
[Music] [Music] for [Applause]
he
[Music] e [Music]
[Music]
[Music] [Music]
a [Music] [Applause] [Music]
[Music]
w [Music]
[Music]
[Music] a [Music] [Music]
[Music] [Applause] [Music]
[Music]
[Music] SC
[Music] m [Music]
[Applause] [Music] hey he he he he [Music] [Applause] [Music] [Applause] [Music]
he [Music]
he
[Music]
[Music]
[Music] track [Music] hey hey hey [Applause] [Music] hey hey hey hey hey [Music]
[Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Music] [Music]
[Music]
[Music] [Applause] [Music] he [Music] he [Music]
[Music]
oh
[Music] oh
[Music] w [Music] [Applause] [Music] [Applause] [Music] [Applause] [Music] I'm just trying to so this okay I to f I'm just TR to give you [Music] something I'm just TR to give you something I do you I'm just TR to you something [Music] w
[Music]
[Music] [Music] I'm just to give something I'm just TR to give you [Music] something I'm just trying to give you something [Music] I I'm just trying to give you something [Music] w [Music] w
[Music]
[Music]
[Music] [Music]
a [Music]
[Music] d [Music] [Applause]
[Music]
[Music] [Music]
[Applause]
[Music]
[Music]
[Music]
[Music] e [Music] the
[Music] I
[Music]
[Music] a [Music]
[Music] [Applause] [Music]
[Music]
[Music]
just uh about uh 5 minutes here hopefully some our folks will trick on mine's on the res was oh yeah anyway the passw mode for just connecting stuff like a keyboard okay well someone who developed Android left that as a capability
[Music] [Music] [Music] I have noide so am I so my cu the documentation does not tell you really yeah it's basically just left I mean there there's a little bit of hand shaking documentation but kind understand how it work but I never read a do so I don't know well if you ever find the documentation for it you let me know cuz I'm still looking [Music] [Applause] [Music]
[Music]
[Music]
all right let's get going you like give me a thumbs up when we're ready yeah okay folks uh let me take care that welcome everybody this is uh a panel on our Proving Ground program here at bsides LV something we started uh as far as I'm aware pretty much one of the first in the industry um this is a uh Mentor track for firsttime speakers at an international you know security conference so um a lot of the folks that come through are actually you know accomplished Security Professionals already they just haven't actually gotten up in front of an audience and uh you know stared them all in their naked eyeh holes um so we are uh excuse me
joined today by uh our current and uh Future Track chairs for The Proving Ground track and a number of our current uh mentees and speakers in the program um excuse me uh uh guy want don't you to go ahead and uh tell folks a little bit about yourself and uh uh my name is guy mcdell I've been one of the track chairs for Proving Ground since 2014 um and when I'm not here I am a senior software engineer at tenal and I have a recurring maker space habit all right and Phil hi can you hear me okay there we go hi everyone my name is Phil young uh I gave my very first talk here at bides
2012 in the mentor program I was one of the mentees and since then I have been a mentor every year um and if you do it enough they just promote you up to the track how it works um although we were joking earlier about how like the less hair you have the better your chances of being a track here so um what I'm not doing this and by the way I love the program I'm a huge supporter of the program I think it's probably one of the best in the industry um when I'm not doing that I am the director of Mainframe pentesting for net spy so yeah uh guy has been leading this track for like the last decade um and uh
you know is we're we're in a right now we're in kind of a two-year handoff period between the two of them uh Phil shadowing guy this year and next year uh gu is going to Shadow Phil and uh we'll see how that goes but I think it's going to you know we're looking to to take this program up to the next level um K can you tell us a little bit about like maybe how this got started what the original impetus was you know what were what the what we're trying to do with this track sure so the program started say 2011 uh it was it was originally run by Mo and and remember her new
handle that I came on about 3 years after after my wife spoke at the program um and we started doing uh run throughs and they went hey that's a good idea would you like to be a chair and I went uh sure um and yeah took it over outright in 2014 um and the entire time our what we've been trying to do is get across how important it is to be able to tell a story a lot of what we do as infoset professionals is try to communicate stories of risk to people who are in decision or in positions of being able to make decisions about that risk and the better we can tell that
story The more likely we are to succeed in helping to protect people and protect the organizations we work for and so being able to to present at a conference like this is one way of getting better at telling those stories and so that's basically the empasis of the program we pair you with somebody who has a lot of experience in speaking and you spend basically 10 to 12 weeks depending on the year uh working intensively with that person to Workshop or talk and at the end you get a speaker credit at an International Conference so yeah what uh what kind of impact uh do you think this program's been having have you seen you know changes in the
industry changes in you know uh uh speak you know people have gone through the program coming back you know what what uh like what's made this worth it in some ways it's it's interesting watching some mentees it's very clearly their they're power leveling through their career and they end up being in positions this guy you know started as a speaker and now is running testing at a on on Main frames um I've seen folks get their work covered in the BBC I've seen people on uh NHK and National Japanese national broadcasting um it it it it basically just it's like a shot in the arm for your career how about you Phil what uh what
have you seen like you know you I know you're involved in other conferences you know very much um so um for those who don't know know um this program has been it's it's it's not just impactful here it it influences other conferences and I know that for a fact because I helped someone put a you know put a proposal together where they created the black hat coaching program that's not a thing that existed like in 2019 or earlier and and that is a direct you know was a direct cause of this program here I was talking how wonderful the program is here and how black hat would benefit from having speaking coaches or anything like that and then now they have a very
robust coaching program it it's run completely different than the program is here but it's the same effect the people are confident in giving their talks because they've had someone who's had that experience coach them through it so that to me is the like like like it's spreading which is you know speaking is hard and speaking for the first time is extra hard and so it's nice to have someone in your corner helping you do that so um what would you say like are and you know this is open to either of you and then after you answer I think maybe we'll we'll introduce some of our panelists here and kind of get their take on this as
well what are some of the most memorable like talks or moments within this program for you one of my favorite talks of all time was actually one of my first mentees who talked about the eth ethical implications of social engineering with robots and one of the things that she did was basically recreate the um not the Stanford Prison Experiment it's the other major psychology test but with robots um and basically getting people to do things that they wouldn't otherwise do just because a robot told them to do it and I was just like that is terrifying and also amazing um another one of my favorite talks was somebody who had gotten a summer internship at NATO talking about how
they pent tested satellites and this was two years before deathcon started their satellite hacking competition and so that was just fascinating like here's how we take this thing that's in orbit that could fall on people and make sure it doesn't fall on people so yeah that um every year we've done Proving Ground pregame has been extremely rewarding because we get to see people who are we we get to help them get over the butterflies we get to help them you know just sort of put that last little bit of Polish in the talking before they send up the dooring that's that's always very rewarding those are the those are the ones I can think of at the moment but
you you definitely stole mine so thanks for that uh one of my mentees um she was doing research Maybe maybe a different talk but also about satellites and adversarial threats to satellites and it was a very cool talk about like and that's nothing look I work at Main frames that is not something that I would get exposed to if it weren't for me being a mentor right like because there's so much happening at the con you know and so I get exposure to all these things that I would never ever get exposed to um couple other memorable ones I know Cheryl she gave a talk on Shadow it that was really really excellent and from
that talk her career i' I've watched her career just explode over the last like eight years or so so yeah go ahead I just I just thought of one um there's a group of researchers at a University of Florida who used zmap to find open core routers core routers that had open SSH they were accepting open SSH connections and they hit all of these routers and basically asked what key exchange algorithms they accepted and discovered that something like 4 40 or 50% of them still accepted deaths oh and when the isps who ran those core routers discovered this all of a sudden the key algorithms changed so that is an example of a direct impact that a Proving Ground
talk had on the state of theart no and what I love about these is it shows like two things that I really love about this program one that this is for first time speakers not necessarily for people who have never like had an extensive career yet right you get people people who are are new out of school but and have a great idea but you also get people who are like accomplished and have done like amazing careers and this is just their first kind of getting that out into the community right uh and then the because of that because we draw from kind of a different you know more varied crowd like the the the topics that we're that
we cover are just fascinating variety right in this one track so and these talks when they come out uh you know they tell people this is not a kidy pool right like by the time these things go to stage they're as good as any talk in the conference it's like I'm very excited there's a talk coming up today that I really want to see it is about using Skyrim to hack other people's computers with mods that is it's such a fascinating thing and I've already seen the top once so I know it's really good but um but that's the kind of diversity of talks like from satellite hacking to mods from a 10-year-old game right like like it's
it's just amazing yeah so in that vein what I'd like to do is kind of go down the line here in the panel and you know tell the audience uh you know who you are you know what what you do how you um got to Proving Ground like why you decided to do this and then what you're talking about or what your talk was about when you did it right hi everybody my name is Le alpi I'm originally from Mexico uh currently working as a security engineer for Google um this is my first time speaking as Vis Las Vegas so I'm very happy about I have been coming for the last three years so I feel like it was time for me
to contribute back to the community so yeah like my talk was yesterday it was a a CBE on open source software very popular that I found last year and with that I did a research uh as somebody mentioned uh we may have like a technical skills but I I want to improve on my soft skill communication skills so that this was a great opportunity for me to pair with my mentor uh who teach me u a lot of things about how to communicate better so I'm very happy and I really appreciate this opportunity all right hello everybody my name is master Chen um I was a Mente for Proving Grounds back in 2014 where I did
a a talk on being a car artist and then uh I've been a mentor ever since um my day job is devops my night job is stalking and ENT um and I want to I to help build the community I want to bring up you know the Next Generation and that's why I'm here I'm lilan Ash Baker I'm a product security engineer at uh the Boeing Company and whisk Arrow um I'm a this is the first time that I've done mentorship here at The Proving Grounds track um my Mente did a presentation on airline Fair hacking and how to read a lot of the data that's available out there within these internal systems um one of the big Parts about doing
mentorship here uh with Proving Grounds track and everything is not only what you can impart and teach others about but also what you get to learn as a mentor about your own talk skills your own speaking style your own presentation uh the way that you put your own presentations together hi I'm caros Gonzalez I'm the CTI leader at Bodo Brazil Brazilian public bank and it's my first time here speaking in in at besides my talk will be about how we are bringing uh TR and tail data to drive the red red blue team purple team exercises and bring changes to the organization much much faster much much more focused
much hi I'm uh Paul Wartman um I'm a Bluetooth security researcher and a research scientist for um Wells uh I'm here as a mentee and give my talk today at 3 p.m. come and see it uh which is on Bluetooth research and trying to improve and augment the security community's ability to kind of explore the Bluetooth Wildlife that's out there and and really just improve the community's ability to dive into this space a lot more easily um I would say that what I've feel like I've gotten out of it so so far though you know just nervous with the call with the uh presentation coming out is um it's been really useful and helping to
kind of reconfigure my thought process on how it is I should present the information that I have how to best kind of reach the point of what I'm trying to tell you without just fire hosing information at you in the hope that you're as curious about the subject as I am hi my name is uh ju I'm a senior engineer from a company called Cloud kitchens uh four months ago I could not have imagined giving a talk at this conference uh and this Proving Ground actually has been fantastic I feel tremendously lucky uh to have a mentor um Australian bloke um to help me uh that's really cool um actually I always want to give a talk somewhere but never
knew the best way of doing it and U my manager just casually said hey why don't you submit a talk to bsid so I worked hard on that and the deadline was actually like Fast approaching so I SED submited literally like the night before and he said basically it's not a security talk if you don't submit the last minute so it made me feel a lot better and I didn't hear back for besides for some time and assume that basically it was going to be you know a wash but I was able to um hear back and actually OB obviously be here so I'm tremendously excited to be here I'm actually very grateful for the
experience and the opportunity to uh to do this thank you thank you I think we have two more here in the in the front would you like to come up to the mic here and tell your story
hi I'm Jen hman I'm a technical director with the government I work for the National Security Agency um this was and this is my first time being a mentor um and it was fantastic I'm OG I guess what they would consider OG right you've been around long enough you're OG so you know if you're in the career field for like 30 some years you get tired of hearing yourself talk so this is a great way to prepare and help new voices in the community come together so and my guy gives his talk um this afternoon on Nicks and Flakes and I am not a devops person um but this was a great way for me to get more exposure to
devops and get him to get his crowd excited about devops so I need to connect him with you later and and finally we have soya aoyama who has a a PR carefully prepared a statement to read for us uh this is one of the things I love most please do come up was one things I love most about this program is we work with folks who all the time who are not native English speakers right and we can still help them participate in this given that it's a lingua franka for this industry you know they can be part of the conversation way they couldn't otherwise uh hello uh I'm soya a um founder and organizer of Biz
Tokyo uh my first presentation was at proving Grant in 2017 I thought uh Proving Ground program was great so I've been involved in uh many activity so far in 201 uh 18 I founded bide Tokyo and enced bide Tokyo speaker to present at Proving Ground in fact a b Tokyo speakers uh become a proving Grand speaker and I have been uh supporting uh them as a mentor uh since last year thank you thank [Applause] you right so um I would love if there's any you know I have questions I can ask the panel but also if there's anyone from the audience who would like to you know ask a question about the program we have a
microphone up here so you know please feel free to just you know line up there and and fire away anybody on either side of the room here but um so I I got to say um you know of this experience you know you You' gone through what um what you say was like the biggest thing that you learned like what was your big takeaway was there anything that you know we'll just again just go down the line or or you know and you know what was the biggest thing you got out of this right um one of the biggest takeaway for me was to put myself in the shoes of the of the people attending my talk like
don't make assumptions like uh you know the way to present information uh make sure to always if somebody's looking at your presentation and get distracted uh and then look back to the screen uh the person should be able to very quickly you know know what's going on right so the way we present the data uh is one of the most important things because you could have the best research the most amazing research that you have done but if you are not able to communicate or express it in a in in a very efficient way uh is the same as uh don't not having anything right so that's one of the of the things how I
learn how to do nice presentation next slides and I feel like that's going to improve my my career and those are things that I can apply in my day-to-day job thank you uh yeah actually to to picky back off of that um I think it's it's great to get out of your own mind uh and into the mind of your audience and so uh the this program has really helped uh with with kind of organizing that and just uh making it a really nice uh place and and way to do that so yeah so I actually learned something very important from Guy the other day which is when you're about to do a presentation go to the bathroom it comes
everybody write [Laughter] down well for me I think it was I it's my first time so I don't have that much experience so anyway but for me I think it was to think about uh how what are you going to present and don't don't just read the books don't tell what's in the books tell your story what what your experience how are you doing things what are what were your results what went wrong how do you manage to to to fix that I think that's a a big uh a great big change to just don't don't just tell what's in the books tell your story so piggy backing off that I would I would agree I think I think the
biggest thing I learned was to really um boil down and be concise about what it is I wanted to get across sure I may have weeks months years of of research and practice that I could just throw up on you of of just a mountain of information but really what was clear was to get that initial hook in so that the folks that really are interested that hopefully will help kind of grow and expand the work that you're trying to share are the ones that get the concept that once once You' got that they'll spend the time they'll dig into it they'll they'll spend the time to really get into the research you're trying to share but that's a lot harder
without presenting it in a really clean and concise way uh I think mine is more about self-discovery and confidence um because giving a speech is obviously a big thing to do at least for me and having a room packed with people to see you give a talk really gives you that feeling that you've done something you know worthwhile I feel it's hard to get that anywhere else another thing too I guess two things um is really like the energy people here I've been so excited about excited about everything and there's so much stuff going on I feel like that's a great Gateway way to like learn more about everything in general and having that energy Propel you forward uh into
the Future No Matter What You Do all right now I'm I'm going to flip this so we stop picking on you first every time but now we we'll start at the other end here so you get to go without any time to think about it um but uh what did you expect coming into this program and what surprised you the most about the experience yeah honestly had like no idea what to expect I went to bsid in Austin the year before and uh it was pretty chill like not nearly as big as year so assume it to be just as chill here uh but the number of people showed up was actually quite overwhelming and
also the quality speaker here is also quite overwhelming like the keyn notes given on both states were fantastic and I was very EXC excited about what happened so that that was really awesome um and also my mentor too I had no idea my mentor was so accomplished in so many fields and that guy is actually toing me on how to give a talk so that was like Michael Jordan coaching you to play basketball I feel not not quite but I mean it was quite quite fortunate to be here thank you yeah so I would I would say I think what I expected to get out of this was to just kind of get better maybe at being
able to speak on the subject or or trying to present my information in a more meaningful way um which I think I got but not in the way that I necessarily expected I think my mentor did a really great job of kind of opening my mind to how I can improve what it is I'm saying how it is I can I can get better at it I think I think the main thing that really surprised me about Proving Grounds is um people here are smart I mean they are really really smart and it is a little nerve-wracking to have my presentation be this late in the game that I've already seen some really brilliant talks
from everyone else improving ground so yeah I'm a I'm a little more nerve-racked than I thought I would be yeah I agree I I agree with you and my talk will be the last one and so but I wasn't I just as I was I didn't have much expectation because I didn't know what to expect so it was great it has been it has been great and my mentor helped me extract much more information from what I I was planning at first and it has been a great experience yeah so I'll say that as a mentor I didn't really have as much expectations as a lot of the speakers here but uh one thing that I did get out
of it was the satisfac action of of seeing somebody that you've talked with you've emailed with communicated uh you know worked through hard to massage a a presentation out of a pile of information and see them just absolutely take off and Excel at providing that presentation to everybody and seeing how satisfied they are with themselves yeah well as a as a mentee um I mean I expected because of the way that Proving Grounds was advertised um I was expecting to work on my presentation skills and the the public speaking aspect of it um what I did not expect was how addicting it would be because um I've been speaking every year since that first year in 2014 and I love it so much
um wonderful and then on the mentor on the mentor side uh just meeting people who have great ideas and their need to share that those great ideas uh it's it's a drug a good one a good one it's good crack yeah I I have the same opinion but I have a quick story for that so when when I get the email that I have been accepted to The Proving Grounds my immediate feeling or reaction was kind of like anxiety right because now now I need to deliver right and this is going to be an International Conference I'm pretty sure there's going to be really cool talks way above my level and then uh I start feeling the uh it's a great
commitment right you have to uh meet with your Mentor like every week and you have to practice ice over and over but uh it's a good feeling because you see your progress you iterate over that you get feedback you addressing next week and as he say he like very addictive right like seeing how your your word gets improving over and over like a little bit every time until you see the final result and people ask you good questions and everybody looks happy so it was a great experience thanks uh Phil guy same question for both of you actually thank you uh so one of the things that surprised me the most when I first started being a mentor
in the program was the friendships I would make with my mentees like I just came in thinking oh I'll Mentor them I've given some talks I'll give them some tips I am still friends with a lot of my mentees I still talk to them we're friends on LinkedIn we we see each other here at bsides it was not that's not what I expected to come out of that this program and it it really is you know you really get to build a network being a mentor of like very smart people so for me it's two things one first of all just watching The mentees Stand and Deliver um anybody can talk for anybody can info dump it takes a lot
of focused time and effort to put together a cogent 25 minute talk and not run out of time so watching mentees do that year after year has has been one of the most rewarding things I think in this program the other one is just watching people grow I mean like watching Master Chen go from giving his first talk to like being a regular repeat mentor and speaker and like becoming an expert in the field has been extremely rewarding um watching uh Wendy Knox everick go from giving a talk on uh very early hacker legal stuff to now she's a ciso at a startup uh and that within the last decade like that career progression just took off like a rocket so watching
that stuff has been extremely rewarding for me uh so back to the other end again here but uh just a very uh kind of classic exercise three things about the program what would you keep what would you start what would you stop
right um I would definitely keep the mentorship program like the people mentoring you those are very high quality people and I definitely will keep them um what I will stop um I don't know I don't think there's anything that I don't like about it um okay but what I will start doing you're perfect telling us we're perfect is a valid answer that's great great yeah it's fantastic I really have nothing negative to say um but what I started is that perhaps um tell us to book the hotels earlier cuz I booked one night here and I messed up rest of the booking so you up having to uh share rumors on my my teammates was
here and he he snores a bit so it's not not fantastic but it was really my fault because I didn't really like figure out the logistics um and also I didn't realize defcom you have a sh out going to deathcon so also mess it up too so I'm actually in like a really far hotel I have to like walk to Defcon so really if you can start that like logistic uh chain a little bit earlier that could really help me uh in the future for new people in the future thank you uh what I would keep yeah I think I think I'd have to to voice my agreement that the mentorship program seems absolutely amazing I I wouldn't change
that for a thing um what I would start is maybe start communic or I don't know if necessarily forcing communication but getting the the mentor mentee communication maybe started earlier cuz while I feel like I was pretty ready um from the get-go I definitely could have used maybe a few more practice runs just a little more time to refine it but I mean at the same time really how much can you polish a turd before you put it up on stage at some point at some point you're just you're just doing busy work um in terms of stuff I would stop I nothing yet but if something comes to mind I will I will definitely let you
know all right all right what to keep I think that the you I say the the mentorship program is great I would I really love the the the the the schedule the the checkpoints because it helped helped me get a sense if I if I'm progressing where I should be now am I good am I ahead am I like behind that was great and what what I would start I don't know maybe as I said having a mentor was too has been so good maybe I had a second one work in Paris to get a second a second view I don't know and and I don't what just you you remove what you stop I
don't know I nothing all good yeah I think you're going to get a lot of people saying not to stop anything within the program because everything works out very well definitely keep the check-ins um because as even as a mentor it's really a nice little checkpoint to say like have you done this this and this checklists are great um the other thing that I would add is on the mentor side is introductions between all of us because many of us just showed up in this room together and found out that we're all mentors uh yes that's that's an excellent point right there is uh uh as part of the mentorship program we are I don't want to say siloed but exactly we
we kind of only learned today that we were all mentors together and so I think it would have been a great uh thing to have these conversations a little bit beforehand and just uh introductions and I think that would be that would be a great thing um there was not a lot about the actual program though that I would necessarily change it really is excellent the way it is or else I wouldn't be uh here doing it for almost as long as Phil um and then just the the last thing is um I know sometime you know we we're asked what our expertise is so that we can be matched up with uh the the proper
uh uh presentation and I know that I know that they do their best best to kind of match you up uh with uh with your expertise um I don't know if there's a a better way to scale that or score that um but sometimes I'm I'm looking at these presentations and I'm like I really really really want to work with this one and I know I don't always get what I want so I don't know that's so so start giving you whatever you want is what exactly it's it's Christmas Christmas is early cool I think uh things to keep I think the Proving Ground uh program as it is is amazing the there was a great
speaker reception party so you have the chance to uh get together and and met people uh I think in terms of uh something to add or something to start will be really cool to have and I don't know if it exists already but kind of like a program of how to transition between being a mentee into a mentor or some followup pass the conference and I think that will be pretty cool and in terms of things to stop I I cannot think of anything to complain right now so sorry I apologize um so same same question just feed it on over and then uh I got another question for the audience in general after that there's a Damon and I have a joke
that we've never found a cfp cfp program we didn't hate um it's the one we have is is it's the least bad we've ever had it's like democracy yeah it's yeah it's it's the worst form of of of program management except yellow the other ones but it's it's better it's it's it is the least bad one we've ever had yes exactly um I would love to do a little bit more coaching on how to submit to a cfp because there are every year there are a handful of talks where you look at the title and you go this could be good and then you look at the abstract and there's nothing and that always breaks my my
heart a little bit cuz like there was an attempt and if if we had provided you with just a little bit more information we could have made something beautiful um yeah that's definitely something I think even for the the broader program that we're looking at uhel submissions just I would love to be I would love just to have a a too many to pick from mhm yeah for for me you know obviously I love the program I'm here because of the program um this is this is this is so amazing hearing all the people having very similar experiences to me it's it's it's it's very heartwarming um so I wouldn't change much I think what I
would change is adding more speaking slots MH right um and and right like putting something together for for because like obviously if you're a first-time speaker it's probably your first time writing a cfp and so we might you might want to put something on the website or something that says like this is how you fill out a cfp here's a talk go watch it for proven ground you know if you have questions email us about the cfp if you don't understand something so we can make ourselves more available um within reason to people first time submitters MH yeah I mean there are examples online and so forth but obviously like it's not yeah but it's an art it's definitely an art like
I know I know John's in the audience here I helped them write a bunch of cfps it's an art yeah all right so I I'm just out of curiosity show of hands in the audience how many of you have either as a a had anything to do with the The Proving Ground program yourselves like uh okay we got like about a dozen easy here yeah okay that's awesome um and then of the of of the other folks how many are here because you're thinking about having something to do with Proving Ground and yep okay couple in the back there all right so um I'm going to invite any of you who who want to answer any of the questions yourselves
that I brought up so far go ahead and grab that that mic right there and and and let or or even come up and use this one because we'd love to hear you know what you because you're in front of the 12 people who raised their hand you're all sitting over there together for some reason it's like you can smell each other sniff each other out it's it's kind of kind of eerie actually but um anyone at
all and if you want you you can if you want to be on camera you can come up here or you can do it from back there if you don't fine to go so um I presented for the first time here uh yesterday so thank you for the opportunity for that um and I think that for me one of the things that I found most valuable was just having somebody else to confirm whether or not my idea even makes sense uh I I was not sure if this was a talk that anybody would want to go to if it was either too esoteric or simply not novel enough um and and having somebody to help me figure out what do I
emphasize and yes this is actually worth speaking about was very helpful validation how do I reach my audience that kind of thing yeah that's that's awesome thank you all right um hi Max your me awesome okay awesome I love it I love it this is great um okay so one more one more oh great yeah yeah please yes
all right uh my name is Jonathan fiser I first presented um at a conference in 2022 I think now I didn't directly go through the program but Phil mentored me so in spirit I came through here and like you said it launched for me and I was able to speak more confidently I was able to write cfps more confidently and then build a community and so it's been very rewarding for me but even more so um from the mentor side of things is my first year being a mentor and being able to get back to other people who want to realize their dream of speaking and coming here specifically to get that start is very rewarding and it um it
helps me learn more about what I'm doing what I could do better get and I learned just as much from my mentee as I think she did for me so um but I do have one question for you guys as well do you have an idea of how many people have directly been impacted by this mentorship knowing that it bleeds out into the industry and it's not just contained to this so is the question the the first order like people who've actually participated in the program as opposed to corly mentored and right so it's been R running at least 12 years yeah so we if we do the basic math let's say 25 talks a year
yeah and then at least 500 and then you know second order thousand who knows right that's that's incalculable really but but definitely yeah we I mean we have recorded you know talks all the way back to 2012 you know a couple dozen a year so yeah so thank you so much for all of your help and making this possible okay so the reason I we're here today is because I love this program this program is deeply personally important to me as a part of what I see this is our one of our core missions at bides is being able to open up this conversation in this community to people who otherwise wouldn't be able to participate right I
mean and maybe not even as a speaker there are people probably I show of H how many of you would have been able to come to this conference if you hadn't been part of this program one out of all of you would two maybe your employer might have sent you or whatever but like that and that's pretty typical like we we in a in a cohort usually 34 of them would not ever be able to attend the conference let alone speak at it if they weren't participating I just think it's phenomenal so I'm on a mission and I want you all to help me with this Mission if you're here today you must Care at least a
little bit about this and I want all of you when you're done with your bsides or even maybe not even when you're done maybe during your bsides go out and find two people that you can get to submit to this program next year okay our our our process begins in January and you know I want I want to get as we're saying I want to fill the rooms I want to have so many people that we we need a bigger room you don't want you to kick out you know a breaking ground and you know but yeah anyway um so why you guys talk a little bit about what the process looks like you know like detail
it you know we've mentioned it in bits but i' love people to know what to expect when they apply so we start planning in January and we open the we call for papers and call for mentors usually February right around usually around Valentine's Day then we leave that open for a month month and a half um once we have once we close that we start looking at all the Hawks and we actually do sort of a two round vetting or two rounds of of cfp approval so the first one we make sure that you actually are a brand new speaker you've never spoken at defcom you've never spoken at black hat never spoken at hack in the Box any Global
Information conference at all and we always disqualify about a quarter of the people who apply um once we have that sort of first round of of review then we start looking at the papers and earns can I just real quick though don't disqualify yourselves right let we'll tell you if you're no you're too big for your for this program but he notice he said major International conferences right right if we have a very short list yeah if you spoke at like your local bsides right to you know it's a one room con that's not that's not a disqualifying factor or even if it's 400 500 people like it's that's true that's true too right so like like don't don't
think oh I spoke at this conference like I said I can't ever submit to proven ground and get the benefits no that's not like let us be the judge yeah sorry did please continue I just so we do that first round of review and then we looking papers in Earnest and we narrow it down to as many speaking slots as we have plus two CU we always we we always like to keep a couple in reserve because inevitably somebody will drop out somebody will get sick somebody will have a family emergency we need to fill those slots and then we we do the at the same time we're looking at the mentors so we look at the papers that we've got
we say okay these mentors have Hardware experience these members have social engineering experience these are policy people who would be a potentially a good fit once we have that batch of mentors and we don't always we we try to rotate through we have a we have sort of a repeating cast of characters we don't try to use everybody every year and so once we've got that that group of mentors then we send them a survey and the mentors actually get some say in whose talk they Mentor so we give them the title and we give them the abstract we don't tell them who it is we say rank these on a scale of 1 to five um uh one
being I have absolutely no idea I'll do this if I have to and five being please please please put me in coach uh once we have that pairing once we have that selection then we pair you off and then it's basically just sort of watching and making sure everything stays on the rails we send out checkpoint emails we make sure things are going well um a few times we've had to address some issues but for the most part it's sort of just you know making sure everything is staying on the rails and everybody is getting what they need and then the day before bides we run a program called Proving Ground pregame which is where we set up The Proving
Ground track first and we get it ready to go and then we invite the speakers to come in and give their talk and sort of do a dry run and then we give them feedback so everybody who's sitting in there gets to give feedback to the talks who are presenting and I'm sure for some of you it's probably one of the most terrifying experiences you've ever had but I I know for a fact that it has had material impact on on a number of talks so that's basically it so there there's only one thing I think I I feel like we may be left out a little bit here which is the stipends right right
so speakers get a used to be 500 bucks now it's what 750 or eight something like that yeah so first time speakers get a STI end to cover their travel and other expenses uh which is something subsidize it subsidize it yeah which is something we're only able to do thanks to some very generous doners yeah and I know and this and we don't do this for anyone like we don't do this for Keynotes we don't do this for like this is a nonprofit conference everybody's here you know out of the the the generosity of their hearts and and their desire to you know grow and impact the community but for these particular folks we do provide a essentially a
scholarship exactly and so yeah that's that that often will make make or break whether or not somebody can come as well yeah all right so we're you know coming up on time here last chance is there anything else any of you would like to add you know to you know any message you want to get out put remember this is going out on the tubes yeah so it's not just to the room uh I want to say thanks to guy for running the program um I was invited to be the co-chair this year and I was a little overwhelmed by what that meant because again I was a huge fan of the program I think it is probably one of the most
most important program that any con I've ever been to so I really want to thank you for doing this I want to thank you for running it I know running it running a track is a thankless job sometimes so I I really appreciate you doing it because this this one like now that i' I've helped them run it it's a lot more work than the other like there's a lot more wrangling cats and you've been helping me with our our breaking ground track and other things for you know years and it's like yeah it's still a step function above it's it's a lot more work but it is definitely so much more rewarding to watch people grow into
their their talks and everything so thank you so so thank you so much for running the round of applause for Guy this is I mean all right with that uh unless anyone else has something to add oh on the end please yeah I want to see the difference between the participant and speaker is huge and anyone who's thinking about doing the program should definitely do it uh it feels like a slide like once you start the Journey it basically like guides you all the way down to become a speaker I feel like that's what this program is really good at doing and that's actually very easy Once you actually make the first step so please
do so thank you yep if you're if you're on the fence going well I don't know if I have if you've ever looked at a technical problem and gone huh that's weird you've got a cfp submission you've got to talk true yeah just apply then you just need to learn to tell the story exactly yeah all right so there you have it Proving Ground um again this is one of the the most core most bsides things like right you know right up there with Pros versus Joe's and the way we you know teach people for cfp I mean CTF like I love that about this conference that we're we're always about bringing in and bringing up and and it's not just you
know going out and being a hot shot so all right thank you folks enjoy the rest of your con
good luck
I'm was I got yeah that's that's what my thought is just the waiting
[Music]
[Music] [Applause] [Music] hey hey hey [Music] [Applause] [Music]
he [Music]
he
[Music] [Music]
[Music] TR [Music] hey hey hey [Applause] [Music]
hey hey hey hey hey hey [Applause] [Music]
[Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Music] [Music]
[Music]
[Music] he [Music] [Applause] [Music] he [Music]
[Music]
breaking round uh this talk is about iot security given by milia weting and Dave Bailey uh a few announcements before we begin um if you guys have any questions after the talk please use the audience microphone situated in front to ask your questions um and please keep your cellones in silent mode so you don't disturb the speakers and have fun thank you welcome everyone as he said thank you for introducing us I'm Amelia weeding and I'm joined here by by my friend Dave Bailey and we're here to talk about some research that we did because we wanted to learn how to hack iot better so welcome to our talk hell zero degrees world so quick overview go through our
intro what's the Internet of Things what are the attack surfaces we're looking at what are some hacker tools and we got some live demos so who am I I'm Amelia weeding I'm a Staff embedded hacker I work on embedded systems throughout the day and I'm a sock goon I'm a badge maker I've made badges for for a few years at Defcon and I do a lot of other stuff so Dave sure so as me said my name is Dave Bailey um by day job I'm a senior staff embedded hacker um I uh volunteer at various events around the de MO area and um I've also worked on a couple badges as well um if you're familiar with the
peoduct project that's the that's my project so milia what's the internet of things well nist has a EXP but I'm not going to read it but you've got consumer iot with televisions mes networks lights speakers security systems home applian ovens ovens I mean I got to preheat my oven on my drive home right that makes sense uh locks garage doors pet feeders water bottles my coffee cup Dave's neighbors washing machine so everything's connected right Dave yep so if if you think about you know things we'll go through an example here I sorry iot cameras you would connect them up to the Internet which then you have firmware and Hardware you see them on your home
computer again whole bunch of firmware software stuff running on them you might be able to see them on your handset so and of course there's a Cloud Server I'm sure that's fine and then the ad servers that come along with it because you know everybody likes to have those on them I'm sure so one of the things that Dave and I like to do is we like to look at the attack surfaces and this one I will read the attack surface describes all the different points where an attacker could get into a system and where they could get data out uh I added that data out part because a lot of times with advanced persistent threats it's about
sitting there and waiting and exfiltrating data when you have the opportunity and with the built-in capabilities of these iot devices today we find that have Wi-Fi Bluetooth GPS glowz sometimes NFC they have new technologies that you've never heard of they have old technologies that you wonder why they're using them has anyone ever used a flipper on a Tesla for some reason they control relays valves sensors cameras uh Jack recers got a pretty good darket Diaries episode where he talks where he has a guest that talks about industrial hacking of like Internet of Things industrially there's also remote configuration data storage push alerts mobile apps and even more different ways that you can interact with your Hardware
I mean you might not know how that device pairs with your phone but it just hooks up to that weird Shady Chinese app yep so we're going to talk now about a couple real world examples and then talk about the research that we did so probably most of you are are familiar with this we like to bring this one up because it happened here in Las Vegas um where you know there was actually someone connected their internet fish fish or connected their fish to the internet and then because it was on rather than a segmented Network it was on their Network they able to Pivot and use that device to get in onto the casino Network which is kind of crazy
yep and it's still a problem today so that brings us to the meat of our talk so we kind of rush through the opening because we have a lot of research that uh we've done together and some demos we want to share with you so why is this called hell zero World well uh where Dave and I hail from it gets to be pretty cold outside and you can see there the outdoor temperature is -2° f there and and I purchased this weather station off of Amazon sight unseen no reviews brand new online and it's iot capable I can control it from my phone I can see what the weather is at home I can see what the weather is outside it's
Internet connected it's got undocumented Bluetooth we found out for the setup and management of the platform it's also got a 433 MHz radio so that it can read those weather sensors that you put outside and it didn't have an FCC ID on it and I'm wondering what the heck is going on here so who saw it yeah anybody noticed the temperatures on the previous slide that's uh makes like Las Vegas look cool yeah it makes it look like it's a it's a what is it a blizzard outside here so it's incorrectly displaying the weather forecast it can't it can't display low temp for some reason we don't know why we dumped the firmware we're still trying
to figure out exactly why it does that but if you'll notice the high of uh 32 in the low of 177 that's kind of funny so when does this happened uh I started playing around with the device I noticed that when I switched it into Celsius the 0 degrees showed up but then it underf flowed to 155 there that you see in the bottom right hand corner and then in the top image you can see that it goes from 32 to 191 so it's obviously doing its math in Celsius and then converting it to Fahrenheit but why does it happen we're going to figure that out later on but um you rely on these devices right you might rely on this for
telling you the temperature of a freezer or to tell you whether or not you have to get home sooner or later because your pipes might freeze would you rely on this for the weather or would you rely on this for security I know I wouldn't I unplugged it as soon as I saw it when the temperature started dipping below zero so what else was here Dave all kinds of fun stuff was in here so the first thing we did was of course open it up because that's what you do with iot devices and we found this interesting chip um so if anybody's familiar with the the tuya brand of of iot devices this is what we found in there um the
CBU module and they um you we able through that find some bunch of data sheets a lot of good information from tuya on just how you connect over over the serial line but they didn't talk anything about other stuff on there they they also randomly decided to get rid of a transistor and just put a zero resistor on there I just wanted to point that out because I thought that was funny so I sent Dave the data sheets yep yeah so we were looking at the data sheet so we were able to get the pin out on this very nicely um thank you to to you they talked a little bit again about their their
protocol over the serial Port so we could start reading a little bit of data what was coming in out of it still didn't make a lot of sense yet on it but we did find a few other projects related to to some toia things and said hm interesting well why did we look at the Port Dave so we of course unmapped it because it's an a network connected device so of course you do that and we found this port undocumented port on 6668 no idea what's going on on that port like I said they're looking through online stuff there's a handful of people looking at potentially what's in that in that Port um but nothing about weather
stations on any of that research so it was kind of interesting yep and that is a typo it does say 6888 because when I took the screenshot I was trying other ports on other two your devices to try and figure it out and that's just the screenshot that made it in here so at this point we had enough data for me to break out the tools on my workbench uh we used that out that we identified to see that there was a uart zero and a uart one and on Art one it was just spitting out all sorts of text uh just plain text debugging messages while it's booting up I mean to the point where we even found some what we
later found out were secrets in there uh we soldered the jumpers of the uart serial pins cuz I didn't have my PC bike kit yet and honestly it's nice to have the wires on it cuz even to this day we can still control it over the uart uh we utilized The Flipper ftdi I found the grounding on The Flipper was a lot more stable than it was on the Ard but they did both work uh and I did use a open source tool called BK 7231 tools to dump the firmware and it successfully act uh pulled the firmware really easily and so um turns out you can get x-rays we were trying to figure out what
was going on with the LED LCD controller on it and unfortunately our X-ray tech missed that one chip that we wanted but we thought we would share some of the pictures we got cuz uh this my soldering job is actually uh the one on the top that looks nice the one that came out of the factories are the ones on the bottom there so um and there's actually some of those wires are just traces under the board that they just said hey screw doing an air wire let's just autoconnect it and then over in the bottom right that's the 433 MHz uh transceiver that can send and receive data and uh up here you can see kind of how the wires were
connected when we opened the board cuz these were our wires they they pulled a couple of the wires off when they were doing the x-rays uh but yeah you can see where our wires are and where their wires are and how much nicer ours are and then even the traces under the board that was pretty cool to see um so then we dump the firmware and Dave goes hey Amelia can I have a copy of that firmware and so I say sure Dave you can have a copy of that firmware and Dave says Hey Amilia I know your Wi-Fi password now because they they manage in in this so this is from the firmware dump they actually
dump a Json file that uh has a bunch of information about the the the firmware on there but it also persistently stores things like the Wi-Fi SSID and password now you know they're not plain text plain text but that's just Bas 64 encoding so which is pretty obvious if you look at it on the on the password field the double equals is like that's immediately how I how I told AIA how I found it is because it's like double equals kind of gives it away as padding in base 64 so it's very much like oh hey I know your password and she's like really and now my iot network is called fake news so uh we took the binaries that
came out uh we dumped them using um binwalk figured out what was there and started going through gedra with it uh one of the fun adventures I had here was a block of IP addresses that I found I'm not sure right now cuz I don't want to dox any of the providers but we found several major providers IP address blocks that allowed us to have direct access to the web servers of over a thousand different customers of that cloud provider to the point where I even found like a cmdb in there by just scanning 024 on an end map uh it was it was pretty trivial to find this stuff and and then of course we found our
passwords in there we found keys for connecting up to the service and more importantly we found the keys encoded in an interesting AES style that Dave's going to explain in a second that allowed us to gain some more um access to the device so that's uh the dump of that uh binw walk there so um Dave says go ahead and net cat to the port yeah so you of course open port what do you do you connect with netcat first thing just to see what's in there also we noticed that the data is coming back in an interesting format so it's like hm so we end up actually dumping it through uh uh through some other uh things and
determined that it's actually just some some asy data coming back and again if you if you think back a few slides where they were talking about the tuya stuff they had some you don't you don't need to go back to it it's way far back to it um but they had some this one it was that one where they talk about these header blocks in there so the 55 AA they're like oh that's interesting cuz we saw that right in in the slides that in the in the data that we were seeing over the network and it's like huh I wonder if that's similar to the to the data that we're seeing and it was but it
wasn't the same because the data on the serial Port was uh kind kind of encapsulated but this was actually a little bit encrypted and so we actually had to go and try and find those keys to try and figure out how to how to decrypt the main block of this uh of this protocol but as you can see in there we managed to to find the keys and decrypt it um and as ailia said what was really interesting when they dumped the keys um if you in that Json file that we showed you one of the things in there um that we didn't dump out in in our slides here is the local key that they Ed to to
encrypt this data but if you look at it you're thinking okay they're using AES we found that in the code they're using ases 1228 um of course they're using ECB mode because well whatever um but we were looking at it and it's like but there's only eight bytes worth of data of key so what do they do do they pad it H how did they generate the AES key no actually what they did is that they have an 8 byte uh AES key that then they take the ask key for each character in the 8 bytes and that's your 16 by key so and then that led to him giving me the proof of concept to write a
console that allows us to just send messages to the device so we just tell it what IP address it is we tell it what kind of command we want to send to the device we hit enter and you see here I'm changing it from Celsius to Fahrenheit from my vs code terminal and then it turns out if you go and you take that extrapolate the data bytes versus the control bytes we figured out what was the alarm what was the clock we started figuring out what was the temperature and and then I figured out a way to uh fuzz it so we ended up fuzzing it and as we were fuzzing it I started getting alerts on my phone and then the
next week Dave was fuzzing it from his house and I was getting alerts on my phone so every time you see the HH there or the LL that's literally sending an iot alert to my cell phone and then you see there completely turned off the backlight at one point the thing started screeching uncontrollably just like it just sounded like a banshee like it knew that I had I had touched it in a bad way yep yeah so we're going to do a couple demos here on on the device so um we're going to try and get the the stream visible up on here um but over the network so we have a Wi-Fi network set
up here and Dave's Running some code to connect to it so I can is it on the network yep we're good so from there I just change it over the network so over Wi-Fi from my laptop to to the device yep you want to do that again so they can they can watch so see how it's in Celsius right now and now it's in Fahrenheit y so that's being controlled by a console over here yeah so I have a so just a console app that I wrote that can uh it sends those encrypted messages over over the network to to the device and I can change a bunch of things on it um then it actually reads back and gets data
back from the device so if I go and change one of the things on here see there I hit that alert button right here yep so so we change if we hit we that's one thing we noticed is if we started hitting but on there we got messages over the network from the device and that's actually what started us to try and decode them and then we realized as we were receiving them that we could send which was really interesting that we could send over the local network to the to the device and be able to to uh to be able to control it but one of the fun ones that we have is uh you know
we're in Vegas we all want a drink right it's got to be 5:00 somewhere yeah it's got to be 5:00 somewhere where why do we make it 5:00 here oh all right go grab your drinks [Applause] everyone and so now that you're scared of the iot we got some recommendations for you segregate your iot networks follow the FBI's regulation or uh recommendations there physically segregate them do not make it a VLAN make it a physical separate Network use strong passwords and securely dispose of your iot devices this is something that we didn't see in the OAS recommendations or the cesa recommendations but if someone went through your garbage and grabbed one of these and went and sold it at a pawn
shop or something there goes your iot network password uh if you're a business track your assets have a documented incident response plan make sure you update update update unless you have a specific samp samung washer then
don't no no so what's next yeah so we're going to continue to try and reverse engineer on this we're still trying to figure out if we can dump the keys without actually having to to solder into it but this this GitHub link is our uh code that we have so far un being able to talk to this TOA device um we think it should work on other ones as well but definitely on these weather stations um that's what we've tested it on so far um and then we're going to keep working to see like I said other Tua devices and we're just going to keep having fun with this with this device and that's my cat
Nicola I needed another picture yep we're good thank you everyone any [Applause] questions thank you great talk uh did you do responsible disclosure to you and what was their response I've had some interface with them I'm curious to see what your response was we got set up as oems with the plain text of we are cyber security researchers who have pinned down to your circuit board and have reverse engineered your platform let us know when you want to hear our findings and we haven't heard anything back but they sent us OEM devices so we're good there well well done well done you guys um so my question is did you reverse the firmware to figure out why
it wouldn't display negative temperatures that was on our list to do to we kind of got distracted with some of this other stuff but we do want to get back to the firmware okay I was just it is it is patched in the latest version so once we got registered as an oem we were able to order a sample of this exact weather station but it turned out that the circuit board is version 1.3 not 1.2 and that one has it patched so we can actually do a diff between the two firmwares now and figure it out we just have day jobs and other things going on in our lives yeah yeah the other but if you want to
help us figure it out we we'll happily post the firmware yeah the other thing with the firmware is that they stripped all the symbols so it's really a pain in the butt to try and debug some of the stuff on it well maybe we should uh strip our keys out and post a copy of the firmware okay all right thank you everyone thank you everyone
[Music] h
[Music]
[Music] [Applause] w [Music] oh [Music] [Applause] [Music] I'm [Music] just I'm just to give you [Music] something I'm just TR to give you something I do I'm just TR to give you something hey [Music] [Applause] [Music] [Music]
[Music] [Music] I'm just I'm just [Music] something I'm just TR to give [Music] something I'm just trying to give you something [Music] w
[Music] a
[Music]
[Music] [Music]
[Music]
[Music]
[Music] [Applause]
oh [Music]
[Music] [Music] oh
[Applause]
where
[Music]
[Music]
[Music] hey besides Las Vegas uh welcome to Alicia and Mon's talk on lowlevel shells uh just a quick reminder please silence your phones and please use the microphone in the center to ask any questions after the speakers are on thank you all the best guys hey guys we'll be talking about lws lowlevel shells any anyone has an idea why it's called LOL you'll find out so I'm Lis this is Manish so we work for secured net I'm a penetration tester Manish is a security consultant and for this project I am a independent researcher but uh my company helped me uh bring this forward to you guys so what's the story basically the story is that I was trying to get a net cat shell
and uh my colleague at the time gave me like a switch I was unaware that the switch was a layer three switch and I'm like well you need an IP right right well that's only an assumption the hypothesis is that layer 2 can be used for a net cat shell but this would need to be post compromise so how exactly is that going to work well we made this meme for you to kind of explain it I'll give a couple seconds so the thought process um we'll need a MAC address and uh all layer 3 traffic is encapsulated in frames which means that um uh you will need an IP but uh what tools already exist that kind of
accommodate this and uh there's a tool by the name of AR ARP exfiltrated that was made by Antonio Basia and uh he was able to exfiltrate ARP requests uh back and forth between uh two nodes and he suspected that this is able to this was uh could be used in a in a C2 framework so in other words it's possible but with art so when life gives you Layer Two make a layer two shell bottom line it's possible to encode information in frames but not through encapsulation so why is this such a big deal well the big deal is that layer three and above is highly monitored layer 2 is not which means that Communications don't actually need
IP or port to work so how does layer 2 work well we have 802.3 80211 which is wireless and uh they work on the same frames or same frames same headers and um all all our payloads all our stuff is going to be stored in the in the payload header of the frame um our broadcasts to all to all to all nodes on the network using the fffff uh address um but other protocols can do the exact same thing there's no difference uh The Ether type is what defines the a protocol and an aana you can register your own ether type which we'll get into the next section ether types why are they so important during Defcon 31 l Richard L
he did a talk about this in his uh talk fantastic world of ether types and where to find them uh his talk was more about how to use ether types and how to structure them to uh communicate with devices in our prototype we will not be using ether types specifically we'll be using them to blend them in to uh make it more covert so here's the payload uh we have, 1500 bytes that we can work with so broadcast uh domain communication so as we all know uh whenever layer we get a broadcast message sent on a layer two it gets sent out to all devices that are connected to that switch so why is this important uh we we are we
were trying to see if we could include netcat Powershell SSH commands in these broadcast domain messages which could be sent out to the other systems that are on the on the broadcast domain so uh let's see what are the requirements and what the data flow looks like for doing this so the requirement as Alisa spoke this is a post compromise tool so we we are going to need initial compromise uh in addition to that we need need Python and C to be available on the compromise system uh binary if you're using anything uh for example netcat if you're using that uh we will need that and a lurer of some sort uh so this is in a nutshell what is
happening the victim is uh send sending commands via Ethernet frames and the victim is responding with ethernet frames back with the response so let's take a look uh detailed look at the flow so yeah the attacker starts off by sending ethernet frames containing the command the victim node then uh decodes the frames ex extracts the command and executes it once the command is executed the victim node then again uses ethernet frames to send uh send the information back to the attack a node so let's take a look at a demo quickly uh for this uh before I jump into the demo I want to show the commands that we are running on the victim node we are running the script in
listen mode and setting the the attacker and session ID and on the attacker system we are running it in Connect mode again setting the same uh attacker and session ID additionally we are providing the MAC address of the victim machine uh and then we'll be running who amander command uh as uh Eli spoke earlier about e types uh for the second uh section of it we'll be using STP as the ethernet type uh the code for that is 8181 let me pull the video
so yeah uh on the right hand side we have the victim screen using Windows which is connected uh in listen mode and then we are using uh the Debian machine on the left on the right uh to connect to the victim we are running who am I Ander command and uh we have wi shark running in the background to capture the packets so as you can see uh the commands that were sent were using loop back and the responses that were received were using broadcast messages so now we'll take a look at the ether type that we spoke about so the only difference between the two commands is we are specifying The Ether type this
time around uh to see what sort of communication uh it it happens so the eer type we using it to blend in so again we'll run the same commands who am I enter yeah so let's take a look at the Y capture we'll Pro uh do the filter for 8181 so as you can see the commands were sent using STP this time around so yeah okay so is broadcast communication required absolutely not we could use pointto Point communication as well the reason why we chose broadcast for this particular example because uh broadcast communication is usually noisier and it gets sometime lost uh and like sometimes the security tools ignore
those so to kind of prove the point that communication not only can like we can tunnel uh layer three and above uh we're going to be tunneling netcat in the next uh demo which the Prototype does uh support uh this is the kind of diagram of what's happening I'm not going to go through it but uh basically we do the exact same thing the attacker sends the ethernet frame the victim receives it it then decodes it it sends it to its own loopb back address where netcat is listening netcat can then run uh that command the response is sent to its own loop back address and then out to the uh out through ether uh
ethernet the exact same opposite happens um uh when the victim is uh is listening and this enables uh two-way communication uh back and forth but this not only works on netcat this works on uh anything so here we have the attacker panel the thing that's highlighted that's where you type in commands for netcat I'll give everyone a couple seconds uh the victim panel here what's highlighted is the actual uh netcat command uh on the right side of the screen that's the that's the tool that's being uh for L2 shell okay we're going to make EDR cry now
okay so on the right side or sorry left side that's where the L2 shell is we're opening up a RDP port or an RDP session to the victim machine now the reason why we're oh sorry the reason why we're doing this now is is to show that uh we have the initial compromise and okay that sorry about this we're having some technical issues okay so we have a EDR solution uh that's on the victim machine and the thread detected that's only netcat that's not L2 shell once an administrator um manually locks this up uh it will actually cut the RDP session uh what basically nothing will be able to communicate to that device uh on any
known protocol so that's what happened here however we have here the L2 shell that was uh initialized earlier we can still communicate to it via
Ethernet okay I'll just close that this is essentially what happened uh frames they do not uh get get to detected and they are EDR IDs they allow it to pass through uh this is all the the loop back traffic that happens on the host and the victim uh it never it never uh appears out of the network so uh what are the limitations of this well obviously we got the broadcast domain and the frame headers they're cut off by the by the routers once they Traverse to different networks but uh we can bypass this with the Nick we can do Vine hopping directed broadcast from routers layer 34 application layer protocols uh out of band Bridges uh these are all ways we
can extand from one land to another uh M MLS networks uh this this works on MLS the same way that it works on any other uh ether type uh MLS circuits can enable that communication uh internal IC networks are very flat it uh it allows um this it it would be a prime target for this so uh we could also do layer to forwarding so uh if there is a victim machine that is compromised we could jump from one neck to the other one uh basically hopping the subnet uh we could also create temporary layer three layer Four Bridges with this uh basically accessing restricted subnets that could uh that could have like data uh data DLP preventions on
them so uh the use case that we can think of is a red teaming scenario uh or like a C2 extension basically where uh we are using this for communication between the compromised host so out of the six uh host that are here we are using uh three of them are not using the L2 shell so they are at risk of getting compromised however the ones that are using L2 shell they would be very hard to detect so now this is kind of the Grim part of the of the presentation uh detect uh unfortunately we will need layer 2 sensors we'll need to span tap the the traffic uh this might work on 80211 it
works on 8023 so just regular ethernet uh the entropy and size of packets they're going to have to be looked at uh the expected ether types for example if you have the configuration test protocol on your network it probably should not be there in the first place uh the only way to really know this is to Baseline things signatures uh good luck have fun it might work uh however this is only to detect common commands that they're unencrypted um Zeke might if it out of the box Zeke is not going to detect it and it's kind of expensive to to utilize um another detection is to inspect the loop back traffic if the tech if our technique is used it's an
adjacent issue but it's good when paired with covert communication um it's not just a network or endpoint problem anymore so take a picture of this uh this is going to be how you detect anomalous traffic via Ethernet um now if the traffic is not anomalous it's a lot harder to detect so so here's the The Good the Bad and the Ugly for detections uh if it's ipv4 uh ARP or IPv6 ether type it's significantly harder to detect so the highlighted uh green that's all the detections with anomalies the above is basically uh IPv6 you can't really tell the difference so defenses vlans that kind of depends if you're only working on one nick uh isolated guest networks I couldn't break
through it through that one so it might be good uh there's certain limitations to it and Cloud environments it depends how Layer Two traffic is uh is implemented on them what has been in the wild um there has been a AP group Platinum that utilized so in their C2 framework however when I looked closer into it it seemed to use IP uh IP is not used in this prototype so it's similar but not quite the same so closing remarks uh this is not a vulnerability uh this is basically frames without structure are still frames um layer 2 detections only work for known attacks DPI unless otherwise configured isn't really looking for this uh wireless devices can communicate on
ethernet Dev to other ethernet devices um Works in line so it's system Network agnostic as long as 802 38021 are supported um and has this research been performed uh I referenced all the people that have done similar research I haven't really found anything quite to this extent uh however there is a references page that will show everyone who uh did similar um tldr it's command control Focus uh ethernet transport agnostic uh there's ether type smuggling um this evades all most layer 3 to 7 IDs without disrupting other network hosts and uh how does everyone feel about Supply chains because uh that could be a problematic if this this is used as a technique million frame question are
adversaries using this I don't know but uh now that the ether cat's out of the bag what's next should vendors do more or is this a free-for-all so here are the references and uh big thank you to bsides Las Vegas and the security community special thank you to Steve Porter Doug lease and Quinn Kramer for helping me and the team make this [Applause] happen questions any questions
sorry sorry which slide can can you come up to the microphone please sorry yeah the command to detect the traffic the TCP dump oh I see okay yeah this is it this is only going to detect anomalous traffic this this won't detect uh frames that are uh properly structured and and basically that are made by uh not amateurs I don't know
so I'm understanding this correct like I said this has to be on the broadcast domain the two the two div your victim and your aggressor have to be on the same uh broadcast domain right not necessarily it depends uh they can be on different networks but the actual communication it it happens on land but if if for example there's like multi niik so you can jump from one network to the other okay this wouldn't work like if it had to go through a switch because it would like you said it would strip the uh so I guess I'm what I'm getting at is um the value in this is it would require uh your your um your aggressor
you would have to have some sort of if you wanted to use this as a remote attack you would have to have some sort of channel to get to it so this is post compromise so the consideration here is you already have compromised the victim machine you are already on the same network as them so it could be something that you use for pivoting for Co communication uh so it's basically sort of uh you could think of it as a C2 extension basically so okay but but both both of them would need to be compromised the victim and the aggressor so uh initially once you have compromised you would be on the same network right so from there you could
you could move on to a different Target who is in a different network and then you still have to be able to bypass the EDR to yes if if that is the case but the thing is l isn't monitored that no no Layer Two isn't but in order to get the initial compromise yes yes yes so that that's that's the only consideration here initial compromise so yeah this is after this is post compromise uh and it's for mostly for Co communication so actually now that I'm thinking about this a little bit where this might be a value is like if you're doing a pentest engagement where you actually get physical access to the facility and you
somehow put a a malicious device yeah that connects directly to the network but then you have remote access via some other yes okay yeah okay cool no not that this count what you guys done this is really cool thank you hello just checking my understanding do you need a raw sockets for this to work or not I'm sorry do you need raw sockets to do this or not uh you so you need admin privileges so this requires pseudo but it does like raw sockets it would work on IP right but this would be your frames so in a way yes okay and have you considered doing this same type of payload but using um setting the protocol to be IP
but just have invalid payload inside the IP packets I didn't quite understand that sorry you if you're if you're making raw frames have you considered um having an IP having something that is labeled as being an IP packet but just with random content so we are not really using IP so it's either broadcast or Mac to Mac okay yeah hey are you going to be open sourcing any of the code you show off today we're considering releasing this we we don't really know uh the impact of this right now but uh we're we're considering it uh raise a hands who would like this release so they can look at this we need a lot more hands than
that all right thank you all right thank you on [Applause] [Music] a [Music] l [Music]
[Music]
[Music] [Music]
[Music] [Applause] [Music]
[Music]
[Music] n [Music]
[Music]
I
[Music] [Applause] [Music]
[Music]
[Music]
[Music]
[Applause] [Music] he he hey he he [Music] [Applause] [Music]
he [Music]
he
[Music]
[Music]
[Music] aack [Music] [Music] hey hey hey [Applause] [Music]
hey hey hey hey hey hey [Applause] [Music]
[Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Applause] [Music]
[Music] [Music] [Music]
[Music] beside Las Vegas uh to the talk on curus the multiheaded embedded hacking tool by Patrick Kylie um Patrick please take it over he grabbing one other connector here event will need this gu so any hardware hackers out there just a couple well this this tool is intended to make it easier to get into that so hopefully we'll end up at the end of this talk with a few more uh if you have any questions raise your hand feel free to interrupt I'd be happy to do it this is a pretty informal talk um it's it's geared around more about demonstrating the tool than uh talking about it but I going to talk about the
process of of building it just so you can uh see that uh you know maybe it'll inspire you to try and uh build something on your own before I did this I only had a moderate knowledge of of how to use Eda tools like key CAD specifically and by the end of it uh I felt like I really knew what I was doing so it's a great way to get into that so who am I uh I'm a principal cult uh principal consultant at mandiant a pentester uh I do a lot of embedded systems testing but also a lot of other generic uh you know Network external web app that sort of testing U mobile
application things like that but my my specialty my passion is embedded systems uh and I'm professionally known for actually breaking those systems that I'm working on that's that's uh what I did and probably the biggest the first one I did was uh about 10 years ago uh I had a BMW don't judge I I like the way it drove uh and I wanted to have an iPod interface in it and I thought it'd be easy I thought it just a little USB connector that'll run to the back of the head unit it turns out it was a lot more complicated than I thought but I decided to tackle it anyway uh you can see in
the middle picture here it has uh some Fiber Optic Cables the orange cable is the one that I had to add and splice into a fiber optic Loop uh that's called the most Loop if for those Automotive people out there that's a basically it's a multimedia Loop designed to pass audio data and I actually had to recode uh everything it turns out the USB interface actually routed itself back to a telematics unit in the trunk and then that spliced into the most Loop and then that most Loop uh moved up to the head unit but I figured it out um and as I was recoding the car to tell it that it actually had that interface I forgot to
hook up a battery maintainer and bricked the car uh so what I had to do is I actually had to boot it back up and finally I figured out what it was it was in transport mode uh which is what they do to uh preserve battery life when it's on a roll on rolloff ship moving across the ocean uh once I figured out how to do that and get it back and after adding a few gray hairs to my inventory uh I got it back and and now I had iPod um and I felt like oh wow I I'd actually bricked it and brought it back uh now for a more expensive brick this was a
few years ago this was during the pandemic uh I had one of these it's a Tesla Model S and I decided I want to make it faster uh when I when the car was sold uh I was the second owner it didn't have ludicrous mode it just had what was called insane mode and I had to get that extra half second of 0 to 60 time so I decided I wanted to figure out how Tesla did that in the service centers uh so I figured it out I have a test bench which I I'm actually going to show JTAG connectivity on um worked through the process and figured out how to add ludicrous mode to it I had to
drop the battery pack replace the contactors replace the fuse and reflash the BMS with different code uh during the the refl the reflash process putting the battery back in the car all worked fine uh it accepted it accept the new battery ID but the cars have this thing called the security Gateway and the security Gateway actually stores the configuration of the car what battery pack you have you know what air conditioning model revision you have uh the drive units every component on the car from all the way up to the door handles have their own little code that goes in this configuration and uh me in my vanity I wanted the little icon for biohazard mode even though I didn't have
the proper climate control units for it I just wanted that little biohazard symbol on my uh AC when I clicked on it and because my car didn't have the hardware for that dur
Related talks
51:51
32:48
33:48
54:33
31:06
20:51