The Good, The Bad & The Ugly In Integrating AI With Cyber Security

Assalamu alaikum. Good morning. I am in the middle of a meeting between three people in Egypt. Everyone is welcome to come in. - Decision review. Decision for big events. If you want to do big events for students or undergraduate students, you can do a project, a major project, or a thesis project. I will be talking about the thesis and how to use AI techniques in cybersecurity. My name is Mohamed Al-Khaisha. I am a professor in the Department of Cybersecurity at the University of Belgium. I am a chapter leader in the security department. I have been working in IT for 13 years. I am a YouTuber and blogger. I have a YouTube channel. articles and training courses for free for the general audience. And there is

also that we are currently doing a 3D focus on the application of the software and the software itself. Okay, why? When we hear the word AI, we put it on the word artificial intelligence. AI is a general term. The idea of the film is to make the computer think like a human. The power of technology is a human. It makes decisions like a human. I will go this way or that way, and do the next thing and do the next thing. If I want to use AI as security, I want to use it as a tech. The software is normal, there is no problem. The activity that comes to me, is it from a spammer or from an IT firm?

Or from a person? So this is what I need to do. The machine learning is a proof of AI. There are other types of machine learning. Machine learning is the idea of integrating a machine that depends on itself in a way. I will integrate it on a certain algorithm. This algorithm will then do the work for me. The reason why I don't have a specific code to do something is because I'm in charge of the system and I'm putting it into a different algorithm. The algorithm will do the work and will give me results. It will give me classification. So the idea of the machine was that I don't want to let people write code at the age of 12 without any time to study. I want

to make it systematic. Even people who have AI inside them, they know how systematic it is. So the idea of the machine is that I have a model or algorithm ready, I will do it, I will do it in the data, it will be done spontaneously, and depending on the data, it will be the first thing that will enter the room. The deep learning is that the algorithm itself learns the mistakes. It detects certain things or classifies them. If it finds a mistake, it learns by itself. It covers itself. When we talk about the media, One of the techniques or one of the categories of AI is that we can design them into three categories. The

first category is called supervised engineering. The second category is unsupervised. And the third category is micro-forced. What does that mean? Everyone has their own specialties and advantages. Supervised means that we give them data. The data is like... Data that I give it, it depends on it, and what? It depends on data. I mean, data, because I build data and I calculate it in the right way, God knows if the malware comes, for example, the features, I call them, if it was the application code, for example, between the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the

malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the

malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the malware, the I don't know what to say about this machine, but it's amazing. The biggest reason is

that I gave it to her, but she didn't necessarily think it was right. She's a machine that depends on data and makes decisions. She gives me a design for data, even this data that was created, I'll insert it again and use it. Reinforcement, if we talk about it a little bit, it is able to learn from itself. Let's talk about any technology we have in life You can use it in a good way, and it is supposed to be used in a good way. It can be used in a bad way, and if you use it in a good way but in a wrong way, then it will cause you to get hurt. So, we are going to talk

about security. Why did I choose the term "Security for AI Cybersecurity" to argue the code that I mentioned earlier? It is to see how AI, in terms of security, can be used in a correct way. First, how can it be used, and then how can it be used. So, the good part of the cyber security mechanism has already been developed. Now, security solutions such as IPS, IDS, WAF, people are working on these solutions. What are they working on? They are applying the machine learning techniques or AI techniques, as we have seen, to work. These are some of the applications that are available We can use machine learning to do malware classification to identify features. As I

said, we are working on malware analysis. The idea of making machine learning apply to malware classification detection, is not a new technology. I would like to take some time to analyze this laboratory Data set, from the malware that we have used some features. The features that define the system, like the classification, the number of the malware, and the number of the malware. Even if the malware is the second or third one, or the third one, or the third one. So, now we have to create a big data set. and that's it It can determine the features that we can use from any software to determine the work from the web or not If it was a "Dos Attack" feature that we can use

to determine the work from the web or not "Dos Attack" or not "Dos Attack" So the machine learning must be a beginner before it is installed People who are professional in their work and have a lot of experience They have the ability to make the machine learn the work from the web This is the right way to start the process of making the first steps. If we look at spam detection, for example, some of the techniques that we used were by solving text. We could see if there are specific words in the text. We could tell them, if there is a hyperlink, the text is a specific thing, we put a specific text, for example, I will show you on Facebook's website, for example, You

are a political scientist. You are a classification. Some people rely on the language itself. They confirm that the language is written in the right way, formal or not formal. For example, an email comes from a Facebook company. They analyze it and determine. Big companies like these, when they send an email, it is in a formal form. On the level of capital letters or small letters, they write on it. This is what they rely on. They determine and consider it a classification. This is the third generation of spam. The second generation of spam is now using a picture to buy things. They started to buy things on OCR. They solved the problem and discovered this picture as spam. They put it in

spam for free. They put it in the internet. The same thing with attacks. The one that works, one of the techniques that I mentioned is Machine Learning. It has a signature for attacks, if it comes to the WASP, like the WASP Top 10 Securitization Access Assets. It has a signature for how the things are in the traffic that can determine this attack, this access. The first type of the technique that I have done is the practice case. I can choose the practice case, the mobile without information. I can track it and explain it to you. The bad part is that we have a subject, a kitchen. In this case, we have a subject kitchen. In this case, we have a subject

kitchen. It is a firmware. I will show you how This is a simple way to explain the network. I am the attacker, and I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is

responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is

responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is

responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack. I am the one who is responsible for the attack This is the 7th stage. If we go to the first stage, the data collection stage, if we use AI techniques, the first stage is in 2017, I think. There was a way to make Facebook commissions for people on social media. For example, you give a profile to a person and he will make a Facebook commission. And he will tell you that he is in a relationship with a certain person.

So this is the process of the crisis that you know, when I first went to the crisis of hacking, I tried to do a lot of things. If there was a tool like this, that was made by the hackers, where they work, hacking people on social media, and I will tell you what people do on social media, you can do a lot of things. It is made by the machine at home. So the link is like this, I will explain the news in detail, and the tool and everything. I can bring someone to talk. I can turn his voice into a voice, but I will be focusing on two people. You may have seen many videos, for

example, this one, where the video is shown, where they put the image of an actor, I don't know, I don't know the problem, but the actor is shown in the video. So this is the original video, using deep-fake technology, through AI techniques, deep learning, to make the facial movements of the natural person appear, and the sound that is given, but they put it on the body of the other person. So this is the attack, this is the private step. Exploitation. There are two research papers, they worked as PhD and master degree. One of them is that you can use a technique, deep learning techniques, you can use deep learning techniques to do something, to do a single possible step. It's very easy.

and you can use this to say it and put it in the paper and the paper is in the description and you can send it to anyone to send it to them if they don't have access to the receipt. Secondly, you know one of the things that challenges the website design or the action is human and robot software for the capture. One of the big solutions is the capture on Google. In 2017, there was a decision in the World Court that allowed the person to pass the catching test. This is a new way of learning. One of the problems that could be there is that the system of AI is the same. But when you build an AI system, you have to

get the data from it, and then you clean it, and put the data in a clean data, so you can transfer it to the training model, and then you can use it for a while, and then you can use it to get the data. Then you can transfer any new data you want to the malware, and you can transfer it to the malware, and it will be classified. If someone can create a vector on this ARP solution, or a data set that is based on the model of the data set The investigation of the attack or the non-attack or the attack of the attacker is a repository of data to destroy the attack of the attacker. The attacker

is a repository of data. The data itself is the same. One second, if you want to get to your code, which you were the one who created the training model, and you can edit it, that's also a problem. If he edits the code to make it work, or if he wants it to work in a way that makes it avoid things that are not active, there is no such thing as a prisoner's activity. Okay, the other thing is... But the problem is that you are not applying the AI technique correctly. This is one of the cases that happened a year ago. I'm sorry. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago.

It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago.

It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago.

It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. It was a year ago. And software needs to be in the box. All of them, as you can see, are not software developers or security experts. So they say that software needs to be in the box. It's 100% true, but it's not. But if you look at

the lives of people who are AI, and AI, what do you get? A lot of money. The reason is that the person himself, when he built his model, The problem of AI systems is the problem of the other systems. Or sometimes it's the problem of costs. The cost is not just the money, but the money for the effort and the number of people who are there to work. They don't have the experience, or the ability to work, and they are busy all the time. All of this leads to problems in the AI system, which leads to the food problems. Sometimes the testing of the AI system is not easy. We have tested many things that affect

the human's ability to take these tests. Some of the AI systems have been affected by this. The EU has been involved in privacy and protection. The EU has been involved in this. Or you can do it like some restrictions. Now, the good people want to learn this thing. Or they want to work on it. They want to work on the project. There is a paper that we will see in the next period. How can you choose people's privacy on your mobile? And you do tracking. And you can install applications without any bench. I mean, I just send you an Android application. I write the code and collect data. I would go to the app and check it out. I would check

it out and check it out. I would check it out and check it out. I do tracking and I also install software and render it. This is not an item that I am enhancing for the fiction of the powerful features. It is actually a paper called Powerful. It is a collection of the techniques of machine learning with signal processing. Even the prediction of mobile applications installed on devices. Let's see how I can work on a project. If someone wants to work on AI, a solution AI system, he can follow the same steps. But before I talk about Android, from Android version 7 and before, there were some permissions that any application could take. An application could take what is called zero permissions.

What is zero permissions? There is nothing more important than the Android itself and the operating system that I want to apply permissions to the processor, or to the computer, or whatever, I take it by default. I can control very consistent from the front cover. When I look at the power, the power, the voltage, and the current that is present in this machine, I can see how much CPUs and memory is there. I can see the file systems that are present there. How much usage is there in the system. How much read/write is there in the system. Also, operating system. The kind of Obedience System, the version of the computer, everyone knows why I chose the Obedience System. Because I am a machine learning technician, so I need to

take some features that are available in a specific solution, which can help me determine if there is a specific application or not. So, any application that your device is working on has a specific behavior. For example, if you go to the site of the glass, Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks

are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks

are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks

are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are the same as side channel attacks. Side channel attacks are and therefore, we can remove the decision-making system to use it in a different way. I don't care about that. So, I'm not saying that it's beneficial. Power, usage, and each application has a

specific behavior, signature, and a specific way of working. In the case of a heading system, an application can run on Android version 6 with a different behavior than version 7. This is one of the features that can be used in the system to create a classification. There is also screen brightness, which is used to limit the screen brightness to 100%. When I turn off the power, the screen brightness will be higher. I can turn it on and turn off the power, but the screen brightness will be lower. This is the features that we have developed. Facebook The first step is to read the profile or make a profile for the power. What does that mean? If I read the power usage, it takes a while. I

bring people from my team, I let them download the application, and they use Facebook. I work with them, I work with Facebook from morning to night. I work to read the power of each application they download. and their profiles I worked on this for a week, so it was a power profile site I see every application as a power in the shop Then I wanted to use the data processing What is data processing? The data that I collected from them for a certain period of time I got the CSV file and solved it I entered the training model, I solved it and it was in the screen Then I entered the training model Then Google

is ready to work with us. We have KVM, C4, and other machines that can work with us. Then I created the training for 40 years. The application itself is the same. I started working on Facebook. Then I started meditation. As I said, the data collection was done by a group of data collectors who collected data from different applications. They chose the applications that were well-known in Jordan, because they were the target of the Jordanian researchers. I looked at Google Apps and found that there are many applications and downloads available in Jordan. This is with the people who are now using the app. I used the third group of features. The second group was about doing something called Power Data Sample.

Power is a signal. People are doing signal processing. I'm doing this search now or the security, and people working with security and people working with security, so I decided to return to the department that works with the specialization of power to study or analyze the power, how to use it. Okay. Then we used the features. We used the features with the signal, and we bought the signal processing platform. So when you want to solve the signal, you can solve it with the time domain or the frequency domain. The work that I told you, I explained it to you, even if I didn't get much into it. I told you, you are one who is familiar with these things. So with the frequency domain,

we used the features together. And with the time domain, I used some of the features that I sent to them to make sure I forgot the output I needed. Of course, in the machine, when you have features that are mostly running in a very large format, the more features you have, the more accurate the output will be. The next step is classification. I trained the classifier and the testers. Then I started testing. As I said, testing is one of the most common applications in the world. I used 3D glasses. I used 3D glasses, HTC, One X, Google Pixel, and Samsung. I started the system on the same day, I mean, I started the software on the same day, I mean, I started the software on

the same day, I mean, I started the software on the same day, I mean, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I

started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I

started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I started the software on the same day, I

started the software on the same day, I found that accuracy is very important in the application you are using. If you were using the application for 10 minutes, you would have According to the third part, the percentage of the application is 95% But when the speed increases, the percentage increases, and the data size increases, and the profile increases. Even If I want to use the people who are working on Android after this, I don't know what to do with them. I can make them an application. This application is very simple. It's not that you have to use it for power and so on. You can use it for whatever. You can say, "Who has the most applications

and uses the CPU and RAM?" "He made me a number to speed up the device and use it for two parts." I can make a software like this. And software is not only for CPUs to work with, it's also for the entire network. Even with this attack, you can only work with this feature. You can only work with this feature.