BSidesNYC 2018 - 03 blockchain

um who have been working on some interesting problems for almost two years now um but uh I'm here today not so much in that capacity uh I also uh work with the Enterprise ethereum Alliance to help understand how we can bring uh ethereum technology into the larger banking institutions and also with Hyper Ledger which is much more broad than just ethereum but also about kind of blockchain and business and whatnot um but that's it let's get started okay so 2017 was interesting um these are nine different cryptocurrencies which I guess are probably too blurry for you to see but I think you can see a general trajectory of what was happening in the marketplace over the last couple months

um great I'm not here to talk about that either okay so uh there's also been some security challenges over the last couple years if you've been following public cryptocurrency um I'm not going to talk about those things either but there's starting to be a lot more really great security talks when it comes to blockchain Smart contracts cryptocurrencies and this whole entire world so but I'm not here to talk about that in fact I'm pretty expressly not allowed to talk about that according to compliance so we're going to talk about we're going to talk about other stuff blockchain why don't you just use a database this is a picture that I keep behind my desk from Andrew Miller who's an amazing

cryptographer and it just says if you if you can't see it in the back it says do you need a blockchain no um this is this is what I you I point to a lot of the times when people walk into my office with their next great blockchain idea and I think it's good to keep us grounded so uh there was there have been some great uh conversations on Twitter recently of people challenging what do you need a blockchain for what is blockchain why do we care why can't you just use a database uh and you know 90 of the time we're right uh or you're right and that's okay but when it comes to information

security uh broadly I think we could do a lot better job of getting engaged and it's really easy to just dismiss stuff kind of like saying well the cloud is just somebody else's computer you know well the blockchain is just everybody's computer um but there's now an entire industry of figuring out how to secure Cloud so maybe we should pay a little more attention so why can't why can't you just use a database well usually you can and sometimes you really could but there's human reasons things like business coordination or regulatory requirements or people not getting along or disincentivization to collaborate uh that are reasons why these centralized Solutions have not taken off an example

of this might be something in the finance world like uh trade Registries so if you have something that you'd like to pledge is collateral for a financing transaction this could be anything from you know some money to an oil tanker uh they would like you'd like to make sure the bank would like to make sure that you haven't pledged that same collateral more than one place because then you're over levered and only one entity can really repossess it right so in the event of default well that sounds a lot like a prevention of double spend problem which is the same thing that we have in a cryptocurrency system so people have looked at could you create

some sort of decentralized trade registry because historically centralized trade Registries that cross Global jurisdictions has simply failed to take off there's no supernational enforcement it's impossible to get people to want to to collaborate and there's really no incentive to do so it it disincentivizes participation and actually is better if you don't tell people because you don't want to pledge it because then you can do the bad action right so in that case uh people are looking at blockchain technology but in other cases centralized Solutions do exist but they're deeply inefficient or they have just outright failed the market right so when we talk about disintermediation that you hear a lot about in business now or especially in

banking um that's really about inefficiencies and wanting to go straight from users to from offers to issuers or you know users to Enterprises and creating like peer-to-peer markets that aren't just person to person but are more B2B great but other times think about centralized solutions that have perhaps not fared so well Equifax uh so if everybody gives their data to one centralized third party we're creating centralized points of failure right and if you look at how a regular blockchain works out in the the public space so something like Bitcoin or ethereum data is incredibly transparent and it's distributed everywhere and that would probably never work for business that's okay instead of when people talk about Enterprise

blockchain what they're talking about is figuring out a way to lay privacy on top of this stuff so that now you're actually creating or reducing the centralized points of failure and creating information boundaries around different organizations so that you're limiting the potential downside of information disclosure that gets a lot more interesting so yeah businesses are really experimenting along this entire Spectrum in some cases what are they talking about is it a distributed database yes and that's okay if there's a single system administrator that can pull the plug uh then it's probably a distributed database but when you talk about different organizations getting together and creating some sort of Consortium where there's really no one person to

call to shut the thing down it's not exactly a distributed database anymore and that introduces a lot of new challenges so problems like besides the data privacy how do you handle compliance or what if there is a data disclosure event what is the governance model who decides who can come in and participate in this thing if it's not just broadly open and public if you're a bad actor or you say we're going to kick Bad actors out of the system we're going to rely on legal recourse we want to fold this in with regular existing regulatory schemas uh what defines a bad actor is that based on terms of use how do you actually hold

a transaction for dispute resolution it's a completely different problem set than what people are talking about when you talk about trustless censorship resistant payments that are Bitcoin or other systems and that's okay it doesn't mean that one is good and one is bad they're trying to solve different problems and you under end up with just a different underlying technology stack so if you understand Bitcoin and then you say why would we use this for business that's stupid just use a database we're kind of just missing the point of what mutualized infrastructure could mean in the future so blockchain quote unquote uh has brought quote unquote new ideas to the c-suite here are a couple of them

I have been asked to explain every single one of these things two people with a c in front of their title over the last two years it is somewhat terrifying but that's okay because much like regular people on the street screaming I love crypto it means that when they have no idea what cartography is um that's okay it means that there's interest in these topics and we actually have an opportunity to bring people in that are interested in our work that we've been doing tirelessly and thanklessly for decades and for some reason we want to just yell at them and tell them to get out and they don't get it so maybe we could do a better job

so let's talk about one of these things in practice here zero knowledge proofs this is an area where my team is doing some interesting work and research so hopefully as a case study it'll help make this all make sense so zero knowledge proofs if you're not familiar are a way to prove something to a third party without disclosing the original inputs right so ZK snarks which is a specific type of implementation of a zero knowledge proof allow verification of the correctness of computations without having to execute them or possibly even knowing what was executed so it's kind of bananas right um but basically what you want to take away from this is that there's like a

whole lot of math like so much math Happening Here right number Theory right so you need to unders in order to understand ZK snarks and I would say that in the whole world there's probably like we're talking about a dozen people who truly understand these things at deep deep death depth and luckily I probably get to work with like a quarter of those people regularly um but you're talking about how to use arithmetic circuits we're using homomorphic uh hiding and encryption uh you're talking about expressing things as polynomials and then figuring out ways to make statements about the curves that those polynomials generate without actually disclosing the exponents that create the polynomials right lots and

lots and lots of math uh we would not want and I'll explain in a second how we do not want Java developers at Banks to have to express business processes all as polynomials um but that's that's not the point the point is it works but it's it's it's what's been lovingly referred to as Moon math okay so in practice uh what what does ZK snarks mean well has anybody here heard of crypto kitties out in the yeah all right lots of love for crypto kitties okay so um cryptokitties on the public ethereum blockchain these are all cryptographically unique Pokemon essentially uh you can breed them you can trade them you can sell them I think

at the last count the highest priced uh crypto Kitty had traded for about a hundred and seventeen thousand dollars um that's because everything is meaningless at this point in life so and that's okay so so cryptokitties don't use DK snarks right now but if they did what you could do is if you were going to trade your ridiculously overpriced cryptokitty to someone else you could disclose a single uh trait that they have without letting them know what the entire Kitty genome holds so where might that pop up in other more perhaps real world situations well let's talk about these business uh leaders who come into my office when I get to Point at the no don't use a

blockchain sign some of the problems that they come up in with are they're real problems right so if somebody walks in and they say listen we're worried about fraudulent account opening and we know that despite our best work at AML and kyc and everything else occasionally a bad actor is able to open a bank account here and we know that this happens at all other banks around the world it's kind of a tragedy of the commons problem we want to share that information but we can't disclose it first of all there's regulatory reasons we can't disclose that second of all we're not going to dump everybody in our database into some third-party database because that's ridiculous

um and finally there's reputational risk to people finding out that this has happened so I'd love to get information from somebody else about what Bad actors have done at their bank but I don't want to share mine it sounds a lot like the generalized problems with privacy right like I'm I really love privacy for myself I just don't want privacy for anybody else so how do how do you approach that as a business problem and then they say can the blockchain fix this for me right because they've heard uh cryptography they've heard privacy they've heard data sharing right and in their head this is this is a problem that maybe works on a blockchain and you have two choices you

either tell them like no a blockchain is really bad for that and especially a public blockchain is exactly the opposite of what you're talking about or we think actually what they're talking about is something that sounds a lot like a zero knowledge proof and in a way if you want to share the information or the outputs of azure knowledge proof amongst other parties who don't know or really trust each other that much or don't want to disclose things to each other maybe what you want is something that looks like a decentralized uh smart contract that acts as a zero knowledge black box that routes this kind of information that no one party has the ability to subvert

right so maybe there actually is an interesting use case there and so that's where we can do a lot better job of listening to people's problems more than trying to solution them immediately or take the solutions that they come up with at face value right so when we talk about mutualized infrastructure what does that mean then if we're going to actually be looking at things like zero knowledge proofs in an Enterprise it's kind of scary so um right now we usually see a lot of trade-offs this is kind of one of those uh here's three things that we'd like or possibly six I don't know I was brainstorming um but choose to choose two vertices and traditionally these are

the trade-offs that we have when building Enterprise infrastructure but when we start thinking about how we could use distributed systems and not just traditional distributed databases but mutualized distributed information can we start if there was really robust privacy on top of that could we start actually solving some of these problems in new and interesting ways maybe we could all right so now we're back to blockchain but it sounds somewhat less stupid this time so blockchain robust decentralized privacy supporting mutualized infrastructure so now we're going to talk about a specific instantiation of that so this is not a product pitch but this is uh the product that I I work on for some of my day

um it is open source it's GPL it's on GitHub so if you're interested please take a look uh so Quorum is a fork of ethereum uh ethereum is uh similar to bitcoin except instead of being a decentralized world currency it's a decentralized world computer it's essentially a distributed virtual machine wherein everyone executes every state transition collectively so you all arrive at the same state of the world it's really cool uh but it has some challenges namely the same ones that you'd find in a system like Bitcoin so you have Anonymous actors with completely visible transaction history that's a problem for banking right so what we want to have is known participants but with confidential data

and in order to do that we've created something that kind of looks like this privacy sandwich I guess so uh and if you're if you're well if you're colorblind the top two colors are orange and green you can't tell them apart but it goes with the various headers here so at the top you have a zero knowledge settlement layer which I'll talk about in a second in the middle you have a shared blockchain that's basically ethereum and in the bottom you have What's called the constellation Network and that's a point-to-point messaging system that allows you to address smart contracts to another participant in the network and it may way very very similar to addressing a pgp email okay so on on

the lower part of the sandwich there the bottom bun um you are transmitting a business logic smart contracts just like in ethereum but what you're doing is creating a separate private State try that tracks the state transitions there completely privately that's great it was the the first version of Quorum that was released that was the entirety of the Privacy solution but it does require that you architect applications in specific ways to ensure that you would have mass conservation or prevent double spend and you know that you couldn't potentially game the system everything was private but having private trustless systems is an extremely non-trivial problem so we were kind of there but we realized that by adding this additional

layer on top which is a bunch of shielded tokens there are tokens just like any other kind of erc20 token in an ethereum system so not proof of generate or proof of work generated tokens but you can create them and burn them at will but they're completely shielded and they're kind of bean counters right so if we can have private business logic that uh exists off the blockchain but is synchronized to the shared blockchain so there's no funny business as to what's happening at that lower layer but then on top of that that private business logic can affect transfers of shielded tokens that have all of the really fun properties of a traditional cryptocurrency system now it's getting

pretty powerful so if you've heard of zcash zcash is the existing real world implementation of ZK snarks it seeks to address some of the fungibility challenges of Bitcoin fungibility in banking is the concept that one dollar in your pocket is as good as any other Dollar in your pocket and a 10 bill in your pocket is as good as 10 Single dollar bills in somebody else's pocket uh you might have heard that maybe every hundred dollar bill in the US has some traces of cocaine on it that's it's apparently a fact I wouldn't know I'm just a friend told me that um but you are not actually held responsible for that the interesting thing about a radically transparent

system like Bitcoin is that it's if it's entirely traceable and it's Anonymous it's Anonymous until it's not Anonymous so the more that individual wallets become de-anonymized and the more that regular people are starting to use Bitcoin and want to integrate it into traditional systems and open coinbase accounts and actually pay their taxes on it the smaller percentage of that system remains kind of uh away from from the regular way settlement right so you end up with a de-anonymized system and if you have a visible transactions but also visible participants now you just have Twitter for your bank account so that's where zcash came in but we said well we don't exactly want to use zcash we have this other challenge so

zsl the zero knowledge settlement layer all those tokens and the sandwich I was showing previously are to Z cash in the same way that blockchain is to bitcoin right so JPMorgan not trading Z cash Quorum uses zero knowledge technology makes sense okay great so let's say that this actually existed um at scale or was being used in production information security could start to look very different traditionally we have thought about securing networks in a perimeter security model maybe you're lucky and you're part of an organization that says imagine everything is already owned how do we limit data disclosure maybe that's on a good day right but in general everything has been about keeping people out and assuming that if you're within

the walls you're safe and imagining a system like this operating not just in uh between untrusted participants but even within an organization where everybody quote unquote trusts each other you could start to imagine information security functioning in a very different way you can also start to imagine a very uh verifiable audit logging such that an organization doesn't have to trust itself such that if regulations change tomorrow and you have to disclose new data that you didn't have to disclose previously you can prove that that data hasn't been manipulated since back then right so different different ways to think about it that aren't just about creating censorship resisting payment systems however integrating this stuff with traditional Enterprise systems is not

well understood nobody has really done this despite the press releases none of this stuff is in production at any kind of a scale that's more than you know double digit numbers moving a day if that right and that's okay you know a lot of it is a hype cycle right now but you have to get the news out there um every you know it's the race to innovate but what we have is a very long road to figure out how to make it real in the meantime so where I'm at now I guess is asking people how do you get involved and assuming that not everyone is going to be doing kind of first principles r d

on these systems but that at some point this will be as uh as as kind of out there as just using uh Cloud native systems you know what would it look like or what kind of questions do we have to answer to integrate these systems into the the ones that we know and love today so from a network security standpoint if you have mutualized infrastructure you know who is granting access how are we handling onboarding and off-boarding how trusted are the other participants depending on your trust model you can use different kinds of consensus mechanisms if everything runs on Enterprise Hardware maybe something like a traditional rafty paxoc kind of thing looks fine if it's not you've got a

million participants in some supply chain use case well your private permission blockchain starts to inherit a lot of the security challenges of a public chain so maybe you start looking at different kind of Dos prevention mechanisms that start to look a lot like either proof of work or cryptocurrency so again it's that spectrum and it's really use case dependent um how do we integrate these things with existing Network management tools how do we create peer-to-peer networks using existing firewall rules anybody who's ever had to make I think at JPM they're called perimeter infrastructure risk assessment Pura review analysis requests uh in order to open a single port on a firewall can understand how creating Dynamic peer-to-peer networks where

nodes come on and drop off all the time could be a challenge right when it comes to appsec it's kind of a nightmare right now we are reimagining all of the challenges of the 90s in brand new terrifying ways but we could be doing better and with better engagement from you know folks in this room and abroad in the space we could be doing better a lot faster so how do we integrate or build libraries for existing scanning tools how can we make formal verification something that's not scary within an Enterprise but is actually welcomed you know just because the language is considered exotic right now does that mean that it's a bad thing how

do we work with our existing csos to get them to expand the sort of of encryption Suites that they're used to using to include something like lip snark like all these things have never been done so it's a pretty great time to try to actually tackle something new and also so many conference circuit talks here I mean you could just do nothing but go talk about solving one of these problems for a year so if if you want to go like thought leader just you could thought leader all over this for like two years so go for it okay so organizations will need people who understand the security nuances between distributed databases distributed ledgers information

blockchains and actual crypto asset blockchains all of which are different architectural things that means that we need you so never forget that actually all mentions of hashtag crypto refer exclusively to cryptoklitis the pleosaur reptile from the middle Jurassic Period of England thank you

I guess how long any questions any thoughts yes go ahead

20 years ago

sure okay so the question was four institutions that still use Cobalt what do you do to ensure that the choices we're making now are not a massive liability uh that is a good question a lot of this stuff is experimental and a lot of the Enterprise blockchains that you see of which I could probably rattle off a different dozen or two dozen of them a lot of them probably won't exist in some time so the race to production does people a disservice right but designing things collectively is a good thing uh the Embrace of Open Source technology has been a good thing um where you know as one specific example of a concrete project JPM is

working with ic3 which is the institute for contracts and cryptocurrencies that does a lot of kind of low-level r d research at Cornell University and elsewhere and one of the problems that we're working with them on is is about integration and the very very long tail of how to Sunset existing systems into a system like this how can you give existing databases blockchainey properties without having to transform your entire infrastructure maybe it's all people want is actually a database with strong signatures and that would be enough um so there's no kind of future ball right or you know crystal ball right now that I can say this is exactly what the stack of the future is going to look like

um and I have worked in in this industry for I guess more than 15 years now which is terrifying I guess um and I have never fully sunsetted any system so I guess the punt to that is probably Cobalt will outlive us all some claps for that yeah

in a blockchain arms race with whom oh other file systems oh Financial Services um banking computers what uh um I think I have never seen uh such a willingness to collaborate within the financial industry uh people understand that these things only work if other people are working on them it's a completely Network effect driven technology so if you work on something like AI or machine learning you can pick a specific product vertical or chat Bots or whatever you know and you can pick a product and you can a B test something on your existing users you don't have to ask anybody about it you can see if there's an Roi you can do whatever when

it comes to blockchain you just can't do that if you're not working with other people the system just doesn't make sense um so it I think it's a massive uh Catalyst for collaboration but everybody of course kind of wants to be there first and I would say thankfully like the ability you know I never really imagined I've worked at JP Morgan much longer than the blockchain team has been in existence I never would have imagined that I would be in a c-suite explaining some of these Concepts to them so just the idea that people are interested shows that they see real business value in these problems [Applause] all right guys so uh guess what we're

only like one or two minutes behind who would have thought uh