Latest / CXOTalk / AI and Cybersecurity: A CTO Perspective
Transcript
- 0:00We're discussing Ai and cybersecurity with one of the
- 0:04foremost AI researchers in the world McCall.
- 0:08Pacu check is the CTO of Jen. Yeah.
- 0:11Thank you know I'm a special breed in a way you know I've
- 0:14been an AI computer scientist for a kind of big part of my
- 0:18life. I've been a professor went
- 0:23through all these postdocs, sabbaticals, getting grounds
- 0:29advising team. 80 students. All this Gizmo and I always was
- 0:34super excited about what a I can have as a positive impact to
- 0:39society. So I was going to one of those
- 0:41crab occasionally in fact researchers and can back in
- 0:472005. We started couple of research
- 0:51projects with my PhD students in the use of AI in the field of
- 0:55cyber security at a time. You know, the AIO Is only coming
- 1:01up as an application area and people were using AI for image
- 1:05analytics. And the video analytics we just
- 1:08wanted to make a breakthrough in the field over how Ai and
- 1:11machine learning can be used in cybersecurity.
- 1:13And believe me those days, it was very difficult to sell AI am
- 1:19L to cyber security Specialists online today.
- 1:22Do they say, I is driving all the cyber security analytics in
- 1:27the majority of the systems that we use.
- 1:29Use these days. And but you're after a couple of
- 1:34startups and to working with the VCS, I was kind of invited that
- 1:39asked, by Andrea Bocelli, who is a former CTO of almost then a
- 1:45CEO. We was being promoted and he, he
- 1:49kept talking to me and kind of getting me, slowly excited about
- 1:52how fun it would be to use a I not for B2B cybersecurity for
- 2:00Enterprise sector but try to use my creativity and experience for
- 2:04building cybersecurity for consumers.
- 2:07Ai and they both are basically for consumers and he planted
- 2:11this bark. And here I am in the gym from a
- 2:16vast Building Systems and running the R&D departments and
- 2:22research labs and threat Labs. So that we build the best in
- 2:26class cyber security for consumers.
- 2:30Consumers are basically, this is what I do.
- 2:34Do you want to give us maybe a brief overview of the
- 2:39distinction between cyber security for the Enterprise and
- 2:44cybersecurity for consumers? This one is extremely exciting,
- 2:52especially these days when the cybersecurity as a field, is
- 2:57undergoing a major change because you're in the the past
- 3:01attackers and the other Us. 30 years, actually at the girls
- 3:06were writing Malabar, and we'll start get it at computer systems
- 3:10at the networks and the programs operating systems that people
- 3:17used it was it was the duty of the Enterprise of the industry,
- 3:22the business to make sure that there is of them, death cell
- 3:26every Hardware, they sell every net with it.
- 3:28Is there is a safe. Okay.
- 3:32So and then users. Consumers were pretty much users
- 3:35of this infrastructure. That the Enterprise sector have
- 3:38made to be safe with the with the recent change that we see in
- 3:43the industry were, it's not any longer, the vulnerability of the
- 3:48operating systems, and the networks and the computers on
- 3:51our devices, by it's more people that are the vulnerability.
- 3:56And the supply chain, people are not only victims of cyber
- 3:59security attacks, but people are also a conduit and Pieces on the
- 4:05supply chain that not only consume but also participate in
- 4:09deploying an attack. So people are getting in the
- 4:13front People Are People cognition and the way how people
- 4:18think and consume the internet is becoming the vulnerability
- 4:22because of this major change there is now a lot more interest
- 4:27in. Consumers are basic rate were
- 4:29the expectation of the industry is to build Old technologies
- 4:35that will be there covering people at the very end or when
- 4:39they touch the internet. And this is a very, and as
- 4:42recognition is what they see what they read the messages,
- 4:47they receive the emotions they post.
- 4:50So the consumer service that created these days needs to be
- 4:54at this very edge of the internet, which is different
- 4:57Edge. That was exciting than 50 years
- 5:00ago. And I'm not only going to
- 5:02talking about about my My what I think but your we have data, you
- 5:08know, Jen as a company is, is a technology company that is a
- 5:16house of a number of Technologic brands in the field of cyber
- 5:21security. I originally come from a vast,
- 5:25which is, which used to be the biggest European cyber security
- 5:29brand that merged last year was not LifeLock, which was the
- 5:33biggest consumer services with the brand and in others.
- 5:39We because of our freemium offering, you know, animals.
- 5:43We were going to do the first and the biggest and the premium
- 5:46and consumer cyber security. We see close to half a billion
- 5:51and points. So we see a huge part of the
- 5:53internet and in this is visibility in gives us a data
- 5:58where we see that. Currently, it's only 30% of the
- 6:02attacks that we see on the internet.
- 6:03And that are caused by a classical model board is
- 6:06targeting devices, and network infrastructure was 70% of all
- 6:12what we see are attacks like fishing and scamming attacks on
- 6:18human cognition. Give in that, weird is a I come
- 6:23into play to help prevent this aspect of the cybersecurity
- 6:30supply chain. As you described it, we see See
- 6:34attackers optimizing their costs and mocked and maximizing the
- 6:38effect of deployment of the attacks.
- 6:41And they were using a different automation techniques and
- 6:44methodologies, including artificial intelligence.
- 6:46Different kinds, not only emotional learning, but also
- 6:50automated planning, automated reasoning different types of a
- 6:53I4 G making detects as cheap as possible and as a large-scale
- 6:58deployment as possible, so there was a lots of AI under the hood
- 7:02or more Writing more attacks mobile deployment, in order to
- 7:06be able to respond and to protect our consumers,
- 7:09efficiently, the cybersecurity firms needed to deploy
- 7:14high-grade, AI to protect the consumers because as soon as you
- 7:22stop deploying the right level of automation against automated
- 7:27attacks, you would be losing your Warfare.
- 7:30There is no way how by even analysts you can defend the
- 7:38automated High scale attacks that are coming from the
- 7:41attackers. So, for this reason, AI has been
- 7:45very well designed to be used on the side of the the defense.
- 7:51But there have been a number of problems that we've experienced
- 7:55as the Defenders. One of those were that for each
- 8:00different type of attack that we see on the internet.
- 8:04We're gonna need it to build a new AI detector, new AI,
- 8:08classifier, we cybersecurity experts were building the
- 8:13classifiers by designing features and doing lots of
- 8:17training. And soon, we started to see that
- 8:22this doesn't scale because we need a lot of programmers, and
- 8:26cyber and subject matter experts who can help us to design those
- 8:31algorithm those algorithms do so much.
- 8:33E-learning tools. You're talking now about
- 8:36signature-based. Yes, I'm talking about the
- 8:39signature bass time over detection, exact.
- 8:42Okay. And in order to be able to kind
- 8:45of cope with this scalable scalability problem, explosion
- 8:49of the types of malware that that we are seeing online.
- 8:54We needed to deploy a deep learning kind of my thoughts.
- 8:58That makes the programmers and the cybersecurity on the list
- 9:03free. Of Designing the features.
- 9:06So there is one type of algorithm.
- 9:08If we're all design that you train on different types of
- 9:12data, large data, and classified the detector exhibits.
- 9:18More General detective detection capability and can be used
- 9:23across the pipeline and the, the cyber security company and,
- 9:29like, in a vast, what we did is we've build those a unique
- 9:33General AI based methods, that help the user has to be to be
- 9:39protected very well and you can imagine.
- 9:41You can imagine this kind of similar to a current way.
- 9:45How AI is effective in building, the large language models, what
- 9:51we did is we've built a similar deep learning methods that were
- 9:55not trained on natural language. But or Json files, internet is
- 10:00written in Json. So like 70% of all Fossil the
- 10:04intent, our Json files structured, but with a variable
- 10:08length, and to be able to train a eye on any type of Json file,
- 10:15I was a complexity that we were trying to resolve and we're a
- 10:18success success success for resolving so.
- 10:21So one Challenge in a cybersecurity is to be able to
- 10:27come up with generic enough AI, that can be effective across
- 10:32different types of current. And future threats.
- 10:35There is the other huge challenge that we have in cyber
- 10:37security. That is explained ability.
- 10:41Yo, cyber Security Experts are like medical doctors.
- 10:45They know the best. They don't need any AI to help
- 10:48them classify moreover, right. So to establish the
- 10:52understanding between the AI researchers and so basically the
- 10:55expert is a non-trivial and ever, which is why there is a
- 11:00need to build a lots of explain ability.
- 11:03It could blow to explain AI to the Marlborough on the list so
- 11:07that it's accepted. So and we were building a
- 11:12exciting novel tools, for example, expandability and
- 11:16cybersecurity, in order to accelerate the deployment of AI
- 11:21in cyber SEC, are the primary challenges here, lying with the
- 11:28ability to accumulate sufficient data for your models.
- 11:33Or in building the broader, General algorithms that will
- 11:39operate, effectively, on those models.
- 11:42It's one of the genetic. Let's say so we have an access
- 11:45to a fantastic data on the internet.
- 11:48So we see a lot, but the generality.
- 11:51So that the detectors are fine-tuned, to be able to detect
- 11:56different types of Marlborough campaigns.
- 11:59This has been a challenge. So as, as much as, In other
- 12:04applications of AI. There has been a push to build
- 12:07an algorithm that is capable of doing more things.
- 12:10Such as in game, playing the designers design algorithm that
- 12:14can play goal and shotgun at the same time.
- 12:17This is the aspect of generality and game playing similar.
- 12:20Aspects of generality is, is needed in building when we build
- 12:25AI for a cybersecurity today, the cybersecurity is different
- 12:29because it's not vulnerability of the operating Eames but is
- 12:34the vulnerability of people condition.
- 12:35So the attacks are different, and the attackers are writing
- 12:39something else. They are not writing Json files,
- 12:41there are not compiling assembler code with the
- 12:44attackers are doing there. There are writing text and
- 12:48natural language. That is supposed to be deceptive
- 12:50and believable so that the users are willing to open an
- 12:55attachment or a willing to share their a cyber say there are
- 12:59financial data there are willing to click on the link.
- 13:03So it's Totally different type of warfare and their artificial
- 13:10intelligence is much more successful and much more
- 13:13impactful but it comes to attacking.
- 13:17Because in order to be able to craft and deploy successful
- 13:21couldn't attack, you need three things.
- 13:24Number one is, you need lots of data about the victim.
- 13:29You need to kind of collect data about where Where the victim
- 13:34goes on the internet. Would they like what they did
- 13:37with the a show? Would they read?
- 13:40And through this? It's possible that the
- 13:43algorithms will create more personalized.
- 13:46Communication, more personalized, could attack.
- 13:49The second is that you need is to steal somebody's credentials.
- 13:52Identity theft are increasing the effectiveness of a couldn't
- 13:57attack. If you receive your the
- 13:59condensate tank from a from an email of a friend.
- 14:03It's More likely that you will click or open the attachment.
- 14:08And the third one is high-performance AI, that is
- 14:12capable of building. A text that is believable, the
- 14:17text that is easy to believe in and adopt as a legitimate
- 14:21message and act accordingly and current high performance.
- 14:26AI is already by the large language models is the ideal
- 14:30tool for an explosion of the coordinate attacks and this is
- 14:35what I'm worried about. This is why I wake up every day
- 14:38and go to work because I want to contribute to protection against
- 14:43AI enabled coordinate attacks. I have been the subject of very
- 14:49targeted phishing attacks. Not have ever been successful to
- 14:53my knowledge but where people have bad actors have
- 14:58impersonated people that I know and in texts and Annie Emails.
- 15:04How can a, i and the tools you're developing protect me,
- 15:10these things are so believable. And I, I'm so used to being
- 15:14attacked that I research everyone and I know how to
- 15:16manually research it. But how can a I help in this on
- 15:20the contrary thinker? I think differently than others.
- 15:24And in the past the whole subdue, the consumer.
- 15:27Subsequently adopted this concept of Cyber security under
- 15:32the hood me as a user. I do not need to understand.
- 15:36I don't need to see. I just buy this product and I'm
- 15:38covered. It's gone.
- 15:40This is history. Now we live in different times.
- 15:43Currently we need to build and gauging sophisticated tools
- 15:47tools that will be there for people will be there with people
- 15:51and will be helping people to be more resilient against
- 15:55coordinate attacks. So, assuming that your I
- 15:58installed this thing, And as a result of this, I will never be
- 16:04attacked by fish. Or scam is just false
- 16:08assumption. This will never happen.
- 16:10So we as a as a Defender need to change the perspective and try
- 16:16to build a companion tools, that will be there with people will
- 16:20be gamifying cybersecurity for people will be rewarding people
- 16:25with more transparency and what is going on when they read and
- 16:30Receive a message. And when you ask me how a I can
- 16:34help the large language models that are now used for creating
- 16:39text. Can be also used successfully
- 16:43for being able to detect text. That is scam.
- 16:47That is fraud that this extortion, that is by many other
- 16:53means malicious. So, the capability of detecting
- 16:56and classifying text Stacks with malicious intent, and this is
- 17:01currently enabled by large animal models.
- 17:04And by AI, which reach, we investigate and study in jenin
- 17:11here are looking for the patterns among very large
- 17:14numbers of phishing attacks and obviously based on your work and
- 17:19your research, those patterns are there if you can only find
- 17:23them quickly enough, I assume. Yeah.
- 17:25And the old, we are lucky because we don't, we don't need
- 17:27to do this work or so this is what the AI does.
- 17:30Us, you know, we only need to provide a good quality training
- 17:34data and then let the Deep neural network to learn its
- 17:41classification. People are always asking me
- 17:44Miguel why I just can I use a GTP to do this for me?
- 17:48And you just just ask the question and the chat window and
- 17:54my response. Always is that cybersecurity is
- 17:57much more serious deed. We Zenda, lots of
- 18:01responsibilities to our users. So whenever we help you sir, we
- 18:05need to be Crystal. Clear, what is it that we are
- 18:08here? What is it?
- 18:08That we are trying to user and if we are advised not to click,
- 18:13we just need to be certain and current setup of the large
- 18:16large, Which models that are generating generally designed
- 18:20Angela trained just do not give you the certain it.
- 18:24So our our added value, and AI pipeline of systems is twofold.
- 18:30Heh. And we just want to make sure
- 18:33that the chat board. The large language model that we
- 18:36query is not giving some of the, your son don't senses answer and
- 18:43second, what we do is we do problem, think.
- 18:46So, we are prompting the language model with the data
- 18:51with like small amount of special sample data that are
- 18:55helping the rational model to do the classification.
- 18:59That is direct contextual to a situation in which the user
- 19:04receives the attack. So we have a question from our
- 19:09Salon Con on Twitter and he wants to know if AI is being
- 19:15used for cybersecurity. Does that mean that Boards of
- 19:19directors don't need to worry about cybersecurity?
- 19:23In other words is the AI just handling this problem and the
- 19:27problem is going to go away. So it's very similar to my own
- 19:31In the third of consumer service regretted, the fact that we are
- 19:34in a world where ordinary consumers need to worry.
- 19:38The same applies to the board members so you can get
- 19:44insurance. But it doesn't mean that the
- 19:46insurance will work always. So the difference between a
- 19:51board is that the bird do have a responsibility for business.
- 19:57So their budgets for an investing in cybersecurity She
- 20:00is a sort of a different order of magnitude while ordinary
- 20:04consumers who payday payday their daily bills.
- 20:09They're they're spotted their Spotify, their Netflix and
- 20:13whatever they need on the internet.
- 20:15For them. Can I think an extra for cyber
- 20:17security as a material part of the bill?
- 20:19So I would say the difference is with the the investment but
- 20:24nobody neither ordinary users. Not import members are currently
- 20:28relief from a Let's intubate the tension and to make the right
- 20:33decisions. These are tough problems.
- 20:36And I think for board members, it's in some cases, even more
- 20:40difficult because they don't have the technology background.
- 20:45That's necessary to really understand this.
- 20:47And so they have to therefore rely upon a group of Technology
- 20:51experts without without the transparency and explain ability
- 20:55that you described earlier. Yes, yes.
- 20:58I would say that. Currently, we have the fastest
- 21:00error, a change in the types of attacks from the Bad actors.
- 21:05So I think that there is a huge expectation for the Cyber
- 21:08Security Experts to really be up to the speed to really try to
- 21:12understand what are the new dangerous threats and also what
- 21:16are the new technologies that people use for attacking as much
- 21:20as for protecting? So there is a huge amount on the
- 21:22experts to really be there for people who need their advice.
- 21:26Many technology experts are Not sufficiently comfortable
- 21:32communicating with the boards, which presents a problem because
- 21:37the security officer, for example wants to explain but
- 21:42doesn't not know how to present it in terms that are
- 21:45straightforward enough for the board to understand.
- 21:49And, and that's a gap that causes problems.
- 21:51In some cases, I agree with you Michael and there's the other
- 21:54problem, which is that there's the current economical
- 21:58environment, the budgets are stretched, Act with with
- 22:01everybody. So and I guess in the past kind
- 22:05of big corporations were, if they didn't understand they,
- 22:09they were going to okay to pay an extra for like an extra tool
- 22:15that this under security guy, requested these days I think is
- 22:20going to be different, you know, they'll be budget fights for
- 22:22everything including cyber security.
- 22:24So this presents a accorded growth for cyber security
- 22:29personnel and Caesars. The big companies to be able to
- 22:32explain better to the board. What is it that they are buying
- 22:37and why they need to invest? So the time is changing for
- 22:40everybody, it is always extraordinary to me the number
- 22:43of companies even security company.
- 22:45As I mean, look at last pass that have breaches and after the
- 22:51breach they always say we're going to invest more.
- 22:54We're going to well, why didn't they do this before?
- 22:57The fact that the old are those centers on the internet that are
- 23:02worth breaching that store users private data and the Bad actors
- 23:07are interested in attacking. I think it's wrong design of the
- 23:11internet. We should have less and less of
- 23:14such places and more and more private data should reside with
- 23:18the users on their hands. Be much more difficult to kind
- 23:22of make a large curved breach, through which you steal hundreds
- 23:28of thousands of of By d's and passport numbers.
- 23:34So I really believe that internet needs to undergo a
- 23:37change, but there is much more opportunity for users to take
- 23:41responsibility for their own private data and not those who
- 23:47just kind of need to validate and verify to check in the
- 23:52Privacy pre-selling manner. What is it that I keep in my
- 23:54wallet and my bullet needs to be secure and modern and good
- 24:00quality. And powered by good compute so
- 24:03that whoever can and needs to check me check my wallet and
- 24:05doesn't need to contain the record of my personal data and
- 24:11the data base, which creates a danger for the vendor, who keeps
- 24:17keeps my data. Since you brought this topic up,
- 24:21can you just briefly give us advice the the the listeners for
- 24:25cxo? Talk are very smart, very bright
- 24:27and can you give us advice on? As individuals what we can do to
- 24:33protect our data. Just along the lines of what you
- 24:35were saying, just briefly very briefly.
- 24:38We going to need to have a good cyber security hygiene to kind
- 24:41of work with a password manager. Do know that do not kind of
- 24:45store or send passwords by a text message.
- 24:50Use do good tools for cybersecurity understand where
- 24:54you weren't, where you are sharing private data for what
- 24:57purpose, whether this is really necessary.
- 24:59Ask Then there's to delete data, because in many countries, there
- 25:03is registration of the vendor, is asked to delete the data.
- 25:07They are obliged to. So be cognizant of tracking.
- 25:13So we attract without tracking. There is no a personalized
- 25:19experience on the internet. So we gonna need tracking but
- 25:22gonna be focused on when. And why delete your cookies?
- 25:27Do not agree with every single cookie pop up that Is bothering
- 25:32you. So these are the basic advice
- 25:35but the truth is and you will actually the storage of private
- 25:40data is is very much connected with algorithmic manipulation.
- 25:46The more data the vendors know about myself, the better
- 25:51personalized digital experience, our receive but the better
- 25:56digital experience means that the the vendors are restricting
- 26:01my choices. When I search for an interesting
- 26:05article if the internet Knows all about me, it tries to
- 26:10second-guess. What is it that I want and
- 26:13serves the content that they assume and I need.
- 26:16So there's there's a piece of manipulation and I truly believe
- 26:23that people need better. People need better tools and
- 26:26Technologies for keeping their privacy in check.
- 26:31But also for understanding how the recommender algorithms
- 26:36That's on the internet. How they work?
- 26:39How they work for me? How they work for me in that
- 26:41situation. In the other situation?
- 26:44I need to understand when YouTube offers me those tiles.
- 26:49Why are there? What is it that I did that?
- 26:52I see this offer what is it that I watch with this added a didn't
- 26:56watch? What is it that I posted about
- 26:58why the recommender is acting in a way he's acting.
- 27:02Well, you know, there's no transparency.
- 27:04We don't know, we are Expect accepting the recommend
- 27:08recommendations from the internet as they are.
- 27:11And this is because this great technology.
- 27:14It's on the other side, the great technology is at the site
- 27:19of the vendors of the intent companies.
- 27:21The amount of technology that is with users in their wallets and
- 27:25their browsers and their phones is actually quite limited.
- 27:28So and as 30 years ago there were kind of first Bad actors
- 27:35and Attackers. And as a result, got going to
- 27:39big, massive cyber security industry, the game became
- 27:43reality, and this industry started to protect users, and as
- 27:47a result of this users are kind of safe.
- 27:51I believe that something similar must happen, in the field of
- 27:55privacy and algorithm on Appalachian that there is a more
- 27:59tank covering people's back when it comes to algorithm
- 28:04manipulation misinformation. Ian and privacy handling, we
- 28:08have a couple of questions now that are popping up on Twitter.
- 28:13So why don't we jump there? And first is from Chris
- 28:17Peterson, who says Ai and ml tool sound great for cyber
- 28:23defense is they're analogous research in penetration testing
- 28:28and Red Team Tools or better traps and honey pots for luring
- 28:35in threat. Hearse in my bread Labs, we have
- 28:39done research and we were able to demonstrate that the use of
- 28:45large language models. For generating malware is
- 28:48possible that you can generate more over by a lot but by GDP.
- 28:56The truth is that is it really necessary?
- 29:01If I look back into how kind of moderate campaigns are Created
- 29:07writing a piece of mother is only a small component.
- 29:10And there are kind of script algorithms for automated model
- 29:14Breitling available in the mower Community for the last 15 years.
- 29:19So is the contribution of a GDP generated Marlborough.
- 29:25So changing for a Bad actors. Honestly, I don't think so.
- 29:30I think that the added value is only limited, however, You talk
- 29:36about the attacks that are in the form of a coordinate, the
- 29:41tank, the form of scan and fishing and manipulation then
- 29:46the story is totally different. There are the role and added
- 29:49value of a large language models and modern, AI or Bad actors.
- 29:56It's just, it's just massive the rate through which they can
- 30:01create a believable unique content is just Amazing.
- 30:07And not only the quality of the content but also the capability
- 30:13to test to A/B test the effectiveness of the cognitive
- 30:17attack, right? You're currently their attackers
- 30:21who are kind of writing fiction? Email they collect some data
- 30:26from the internet to learn how affected they have been.
- 30:29They take some learnings they end up the strategy.
- 30:32They try something else through a method.
- 30:36Two reinforcement learning together with large language
- 30:41model based tax generation. This cycle can get automated.
- 30:46So this is my worry, you're in the Bad actors.
- 30:49Start to really automate to generate a unique dangerous
- 30:55content at the same time to be able to learn automatically how
- 30:59effective data has been and adapt the the text generation to
- 31:05me. This is very dangerous and this
- 31:07is this is an area where we as Defenders need to pay big
- 31:12attention. It's so interesting that you
- 31:15just Tribe this A/B Testing because among the attacks that
- 31:21have been against me because of who we interview, it's EXO too
- 31:25awkward. You know, we're a Target and so
- 31:29I get these requests very, very strange.
- 31:33I got these requests from what appear to be women on the
- 31:39internet, you know, accomplished women and what I've noticed and
- 31:43I'm married. I have no interest.
- 31:45Okay. But I've noticed that the
- 31:48attackers because they're obviously fake because I
- 31:51research everything. So I noticed that the attackers
- 31:55have been changing specific variables.
- 31:59So they'll change, for example, a little bit about the
- 32:03background, all the variables will be the same, the
- 32:05characteristics of the proposed connection, they'll change
- 32:10ethnicity, they'll change the tone a little bit but keep
- 32:15Everything else, the same. And so I've suspected very
- 32:18strongly that that this is a be testing going on.
- 32:22Yes, yes. And this would be testing can be
- 32:24automated if there's a Tracker in your email or tracker in your
- 32:29browser, that is hoping the attacker to can.
- 32:32I report to understand how effective it has been, is going
- 32:35to be automated. We have another question now
- 32:38from Twitter and this is again from our Salon Khan who says,
- 32:44can a I make recommendations about not collecting certain
- 32:51kinds of data since it's prone to be attacked.
- 32:54I think he's referring to recommendations to individuals
- 33:00about what kind of tracking to allow.
- 33:03For example or not based on recommendations from an AI.
- 33:07This is a big challenge or one of my teams we are trying to
- 33:12understand to which extent a I can help.
- 33:15People to make the right privacy decision.
- 33:18So we build good quality AI. That helps the user to assess
- 33:23and explain, what is the Privacy impact of using this or the
- 33:30other app or being on this or the other webpage.
- 33:35So kind of rebuild classifiers and detectors that help users to
- 33:40give them some basic information about that.
- 33:43The truth is, Currently users are not ready for and I used in
- 33:49this information so that they would be impacting their privacy
- 33:54Behavior. So you're we see that users by
- 33:59big big parts are taking a binary decision.
- 34:03Okay. So I don't care or I just got
- 34:07think Anita or I don't care, I use Google search or I care and
- 34:11I use DuckDuckGo. Right.
- 34:12There's actually couldn't nothing in between and I To the
- 34:15technologists and the technology firms need to build tools that
- 34:19would allow users to kind of set up their privacy approach in in
- 34:25very fine-tuned way in the context that the time ago.
- 34:29Based on what they do, where is with a search for base when they
- 34:35are physically and let the user to kind of fine, tune the
- 34:38preferences, and then the able to be with the user and adapt
- 34:43the privacy. Preferences based on the past
- 34:47Behavior models. So I actually think this is a
- 34:50like a missing piece that the users needs to get the idea.
- 34:54That's an optimization problem. To get the best of the
- 34:59personalized, dental experience on one hand and protect your
- 35:03privacy. On the other hand because these
- 35:05are gonna joint. John, bardos macaw.
- 35:09Can you describe to us some of the most significant challenges?
- 35:15Especially the technology challenges that you face in your
- 35:20work and your research right now.
- 35:22One of the really fascinating research challenge is explain
- 35:26ability in the field of cyber security.
- 35:30So Kara for the last 10 years, many areas and have been working
- 35:35in the fourth explainable at trying to deliver some good code
- 35:39explanations on AI verdict in cybersecurity.
- 35:43This is, this is Is this fascinating?
- 35:46Because if you're gonna dig deep into the mind of threat,
- 35:50researcher threat analysts, there is a combination of deep
- 35:55knowledge, great intellect, fantastic reasoning, and
- 35:59intuition. And the intuition is the piece
- 36:02which is very difficult to optimize.
- 36:05So whenever there are AI scientists actually experts,
- 36:08they just want to do statistic, they want to optimize your, give
- 36:12me optimization functional training across the fire.
- 36:16This is, this is, this is how I people work and to be able to
- 36:21marry this statistical approach to world with the intuition that
- 36:28is, inside of the security. That's fast rate.
- 36:31And actually, I think it's transferable it's transferable
- 36:34to to other domains. I would say that a medical
- 36:38doctors are very similar. I think that at a time when a I
- 36:42will be taking over in the healthcare, they would need to
- 36:47be able to resolve similar challenges, the challenges of
- 36:51explaining AI in such a way that we are getting a professionals
- 36:56that are not being replaced by AI, but their impact is being
- 37:01Then X multiplied by a proper use of high-performance Ai and
- 37:06I'm lucky to be in this field that I have a first-hand
- 37:09technical experience. So this this definitely is is is
- 37:14is one of the challenges and then secondly, oh I'm I'm really
- 37:19excited to be in this time of explosion of the large language
- 37:25models and to see the first time in my lifetime when AI is Being
- 37:31a true consumer proposition. Until now a, I actually was a
- 37:36preposition was a B2B Enterprises firms.
- 37:42Corporations were using good quality AI to deliver products
- 37:45to users, but this is the first time where users are exposed to
- 37:50Ai. And I would say that one of the
- 37:54big challenges for me and for my team's is to understand how this
- 37:58new AI Is threatening people. What are the dangers and limite
- 38:05an ethical limitations of this new age.
- 38:08AI that can hack the can have impact on our users and going to
- 38:14be there and to think about those dangers and to try to
- 38:19build a technology that is protecting users against not
- 38:22current, but future AI dangerous.
- 38:26It's very rewarding. We have a question from
- 38:28LinkedIn. This is from From Nash in blue
- 38:31and she is a, she runs an organization.
- 38:35That works with Chief Information officers.
- 38:37And she says this, she says there are quite a few, a I
- 38:42powered cyber security tools out there.
- 38:45What's your take on the level of maturity of these tools?
- 38:50What's the confidence level of organizations as far as adoption
- 38:55goes? Especially since the popularity
- 38:58of chat GPT? Should be explained ability.
- 39:02There are tools that people believe because they've been
- 39:06well explained and the users have trust in the tools and then
- 39:11there are going to New Kids on the Block that have hard times
- 39:14to prove their value. So um, Optimist, there are lots
- 39:18of great AI tools. That's understood.
- 39:21Experts are using, is non-trivial to be able to detect
- 39:25what is working for your use case and was not what advice do
- 39:29you have? For Enterprise Chief information
- 39:33security officers. I know it's a hard, it's a
- 39:36really hard question to ask. Just brought it Vice, but you
- 39:38have such an overview not to play catch-up, but think for the
- 39:43future, I know that cisos often are busy with solving the crisis
- 39:50and with extinguishing Fires at this is pretty much a work.
- 39:54Not try to find at least 50% of your tongue.
- 39:58Where you think about the future about What may come up as soon
- 40:02as the technology becomes better developed, and can more more
- 40:07powerful, do not us as a cyber Security.
- 40:11Experts stuck in the, in the current situation, but let us
- 40:16find time to focus on the future because this is the only path to
- 40:20resilience and any advice for Chief Information, officers
- 40:25Chief Information officers need to partner well with see Usos
- 40:30and cdos and to gonna be in a good company and try to take the
- 40:35advantage of what the Cyber Security Experts can do for them
- 40:40and what's the technologist can help them and driving their
- 40:43future decisions until I don't invite and view the same with
- 40:48investment in security, but will be under scrutiny.
- 40:52The same time, we will see investment in it and clothespin
- 40:57to undergo, big scrutiny in 2023 and R2 a year working on certain
- 41:03problems challenges in cybersecurity and AI.
- 41:07Where do you see the results ending up of your work over the
- 41:12next couple of years? The way how I see evolution of
- 41:16scan and fish the coordinate attacks on the Internet.
- 41:20It's from like short. Thanks to long-term persistent
- 41:27coordinate attacks from manipulation to misinformation.
- 41:31And I actually think that in the future as current, the
- 41:34cybersecurity is changing from attacking people devices to
- 41:38people. Cognition, I think in the
- 41:40future, the subjects are basically will be changing from
- 41:43a immediate kind of attacks, which is this one click towards
- 41:47manipulation. How am I changing changing
- 41:50People mind that? There would be less.
- 41:53Resilient and more susceptible to to an attack.
- 41:58How do I move people between a coach a, so that they would be
- 42:02more vulnerable? They will have more
- 42:05vulnerabilities for me as a bad bad actor.
- 42:08So I think that this combination of a scam and phishing attacks
- 42:15and misinformation and manipulation will be a big Topic
- 42:19in AI cybersecurity in the years to come.
- 42:23So your Saying that the shift takes place from AI in
- 42:30technology attacks. Meaning attacking firewall, for
- 42:33example, to cognitive attack, such as fishing to broader
- 42:40overtime attacks that manifests as misinformation disinformation
- 42:47and broader psychological manipulation.
- 42:50Yes, yes, exactly. And this is where I see the job
- 42:54of a, I substitute cybersecurity expert needs to be even more
- 42:59exciting and would require a wider scope of knowledge
- 43:03comparing to subsidy Experts of the past.
- 43:06It's fascinating, because essentially what you're saying
- 43:09is that the field of misinformation and
- 43:13disinformation, which right now we look at as a social media
- 43:17problem. In fact, it's a cybersecurity
- 43:20problem. It is, it is a security problem.
- 43:23Live on to today because the most effective weapon in this
- 43:29future cyber security, world is making people more resilient the
- 43:33helping people to be fit. Mentally third, to be
- 43:38inquisitive to be excited about checking sources and to be a
- 43:43resilient against future attacks, which is why those who
- 43:48can expect the technology firms to build a manipulation firewall
- 43:52that I put in. My browser and doesn't show any
- 43:55fake news at the wrong approach for one.
- 43:58It it cannot be done and for a second, it reduces my mental
- 44:03Fitness are going to need to be there still.
- 44:05I just want to be on the, I want to be independent autonomous and
- 44:09keep my life under control and to this, also belongs to my
- 44:14capability to distinguish, which news I believe and which use?
- 44:18I don't believe, I just don't want this right to be taken away
- 44:21from it. Seems like this is a A
- 44:23particularly difficult challenge because now what you're asking
- 44:29to take, place is silos of researchers to converge.
- 44:34Because right now, you have cyber security researchers.
- 44:37And then you have folks who are looking at news.
- 44:40Essentially media researchers, these are different groups of
- 44:45people is they are, but they'll be soon.
- 44:48Sharing the same objective. We make the internet a safer
- 44:51place for everybody. Well, Well, on that, I'm afraid,
- 44:55we are out of time so I shall. I just want to say a huge.
- 45:01Thank you for taking the time to be with us today.
- 45:04Thank you for having me Michael. Thank you so much to McCall.
- 45:08Piccu check the CTO of Jan. Now before you go, please
- 45:13subscribe to our YouTube channel.
- 45:16Hit the Subscribe button at the top of our website so we can
- 45:19send you our great newsletter.