Integrating Artificial Intelligence Into Cyber Security

okay thank you my name is of Iran Sofia I'm researcher at ICRC and also I work for a company called Sherpa networks in the network security space in a way my session is an elaboration to the comment made by Professor Bennett's file earlier today about AI versus AI I would like to speak about implementations of AI or practical implementations of AI in the cybersecurity world mentioned some terminology and and conclude with some maybe an insight into the future so let me start with the following observation which I'm sure all of you will agree with we live in an era of digital transformation data is the most valuable asset that we have according to the economists and there are a lot of statistics supporting this claim one one that I found interesting was that 90% of the data that we have today I was actually obtained in the past two years that's an outside outstanding number to make use of that data we need to analyze it it is called a big data analytics and it is all about gaining some knowledge from those large amounts of data and doing it in the context of the business to serve a purpose or to ask a question which is relevant to the business for my presentation the business would be cyber security defense or cyber security offense depending on the perspective so here's the question can we do something practical in cyber security defense with AI or who did that be used by attackers to to achieve their goals obviously the answer is yes and I'll give you some some examples in the following slides before I do that I'd like to ask a question obviously I will answer it myself and the question is why do we see this growth in using analytics now and in my opinion there are two answers to this question the first one would be that the technology is affordable and available just imagine what anyone would have to know to do in order to obtain the amount of computing infrastructure required to analyze big data in the past by everything set it up deploy it etc is costly and takes time versus the ability today just to consume distributed computing as a service from from cloud players like Amazon or Microsoft for example and the second answer has to do with the the progress of of academic research specifically on those technologies analytical technologies such as AI machine learning deep learning that I will mention in a bit and that progress as you can actually feel that progress when looking at personal assistant engines such as Apple Siri or or Amazon Alexa D the accuracy the amount of progress that could be seen on those tools is is actually showing how the the whole the whole industry progressed in Academy and the industry so very briefly just mentioning the buzz words around those analytic texts on technologies so you might hear three buzz words AI deep learning and machine learning you could see on the left that they are sorted by the level of complexity in intelligence they have so AI is the most intelligence ones obviously more difficult to obtain and machine learning is more basic technology less complex easier deploy today what I would like is really mentioned machine learning is the notion of machines learning by themselves deep learning is kind of a manipulation of machine learning imitating the way a human brain works and artificial intelligence is really the highest level and this is machines doing what we call or consider intelligent there are all computer science technologies and you could see that they also subsets of each other so what is needed to do this right basically there are three areas that need to collide or to exist together to make a data science properly and one of them is get the data right use coding and computer science or hacking sometimes it's called to get their data right there's a saying called garbage in garbage out which means if you don't get their data right your results are not gonna be correct the second one is do the analysis properly for that you need pH D you need the math or statistics or stats expert to do that and the third aspect is domain expertise so if we are asking questions about cybersecurity we might as well have a cybersecurity expert within the team between this to make this engine so he can ask the right questions and also interpret the answers understand what the answer really means so let's take a look at the two implementations of of AI or machine learning to be more basic in the cybersecurity world one of them would be for detecting malware in in files so you might also be aware that it's an ongoing arms race between attackers and defenders the attackers hide or better hide malware in files every day and the defenders struggled to find whether those files are malicious or not the latest hottest technology is using machine learning to do that and in a way it is like a face recognition so the way face recognition roughly works is that there are features in the face that help machines identify faces and files also have features inside them they have there are there are a lot of ways to exploit those features from files and machine learning technologies can do so and then determine with the high probability whether the file is malicious or not so that's the first implementation second one is really interesting as you all know the web traffic is becoming more and more encrypted basically it is used to announce security and ask confidentiality privacy but that makes it very difficult to any security engine to understand what is going on with that traffic so just just to share some statistics recent google statistics have read says that 81 of the most popular websites in the world from the most hundred popular web websites in the world is encrypted by default Google plans to announce on their browser every unencrypted web site is unsecure so they will change the definitions of the browser so we are really going towards a fully encrypted Internet Society and also as professor Bannister mentioned there are ways using machine learning to understand whether file or traffic is malicious or not without decrypting it without looking what's really inside he mentioned that in the context of keeping the privacy but it also are very important in the context of keeping performance because decrypting encrypting is really a painful task to any machine let's have a quick look at the other side the other side is all about one manipulating machines and the example I've taken here is from another discipline or from another world but it is the same principle or it could be done in the same principle in in the cyber world it is about using the rules of the machine against itself in this case this is an autonomous autonomous car it was tricked to get into this circle as you can see the rock the signs of the road allow it to get in but they don't allow it to get out and there's no the machine doesn't understand the context it doesn't understand that the parking lot is fully empty and available every 5 years old kid would understand that but the Machine doesn't so this is in a way this is a black box attack this is without any manipulation without penetrating the machine it was actually tricked so same principles could apply also to different security engines the other example I have is is about fooling people not fooling machines so as you all know social engineering or manipulating the art of manipulating people is is an important and effective attack vector or at least a way to to obtain a targets so what we see according to researchers and obviously also in tools is that machine learning is is starting to play a role here in obtaining targets in finding the right people with the right amount of money with the right information about them so any attack could manipulate act could manipulate them better and this is will probably be always the the weakest point the human beings so let me have a quick look into the future and ask the following question so we've seen AI used as a or subsets of AI uses a technology for defense and it could be used as often so and the question is can it be more than just the technology in the cyber world and I'll let you know or I'll share what I mean if we if we take the analogy of the car the autonomous car we all know that since count Benz introduced is the first car in 1886 for let's say a hundred and twenty something years basically nothing much has happened so the car the engine is faster and the windows are nice and the quiet the car is quieter but there are still there's still a wheel there's a driver in to look everywhere he has a brakes yes a gas nothing really happened until this this came in so this is a disruptive technology autonomous car changed the world in terms of automobiles so the question is can a I'll do the same for cybersecurity can it be more than just a defense or attack technology so let me offer a possible future obviously I don't have my crystal ball but it is a possible future and the future is called the the self-driving Network it is all about the infrastructure that we use the digital infrastructure that we use computing infrastructure that we use acting within an intelligent entity AI entity to defend itself to recover from failures to automatically adjust resources etc so without getting into too many technical details there are some technologies today in the industry that are starting to to move towards this direction and this is a possible future I think it's also a very very or could be a risky future and many people ask the question do we really want the machines to take over do we really want AI to become so dominant in in any area but let's talk about cyber security as an example and I'd like to present two points of views and with that to conclude there are obviously opposite to each other one of them is the point of view of Andrew a professor Andrew a bank he's the he's a former professor in Stanford University is also the co-founder of of Coursera the online Academy website an institution he basically says AI will be like electricity so when he says electricity means it's going to be everywhere it's going to be parent to us so we don't really feel that we don't pay attention it's all over us and it's going to be benign so it's not going to be harmful it's going to be the new electricity so that is one point of view many people share that point of view but there are others really serious people like Ian masks for example that really see in AI the demon I don't know if you heard or read some quotes from Elon Musk he actually called AI he actually said AI is worse than nukes he called AI the demon and there are theories like the C the technological singularity point theory that claimed that in a possible future a machine will run into a surf learning mode that will eventually make this machine smarter than humans and beyond that point nobody knows what will happen so I don't know what's going to happen but those are the two points of views and we are looking forward as as the industry and academics and also as a society to see what will bring us the future and with that I'd like to thank you for your time for the opportunity to speak here for staying late thank you so much thank you very much of you and well thank you for all the speakers for this mind-boggling day I think we heard so many new ideas and approaches we understand I think we can conclude two things from this first of all we are in the future already we are living in the future and the second is that we are not hopefully middle man not the man in the middle I hope we all know that there is huge distinction between these two we are a middle man in this world of try to connect technology to our to our jobs as representatives of our countries we connect people and what we do here is the same we connect the technology to our challenges and you heard from different angles how this works we are here because you want to protect our societies and we want to protect democracy and you want to maintain the way we live we've lived in the past you want to continue that in the future but we need to understand what technology really brings to us and so with these brilliant ideas that researchers that are all practitioners from the tel aviv university brought to us i like to leave you with the idea that the upcoming conference in June will be this but much bigger and will be all looking very much forward to having your contribution and your representatives join us there and looking forward to sharing our ideas there and food to the discussions there so thank you very much and say and have a great day [Applause]

Channel: TAUVOD

Views: 2,708

Rating: 5 out of 5

Keywords: אוניברסיטת תל אביב, The 4th Ambassadors Summit, Digital Diplomacy, Blavatnik Interdisciplinary Cyber Research Center, Tel Aviv University, Cyber Security

Id: 0GQKFZz_UCA

Length: 16min 42sec (1002 seconds)

Published: Sun Mar 25 2018