Computer Security in the Age of AI – San Francisco 2025

So I'm going to speak a little bit about uh computer security in the age of AI. I'll begin with uh two examples. The first one is um a case where uh cyber criminals were able to use an internet connected fish tank uh that was present in a pretty large casino in the Vegas uh to excfiltrate data out of it. they were able to use the fish tank. Uh they were able to exploit an issue in it and they were using it to take the data out of that casino.

U that's one example. Uh the good thing was uh the casino was using um cyber security technologies that had AI built in which was able to detect this unusual traffic and they were able to uh friend this attack before it could really become uh catastrophic. Uh the second example that I have is this one where um an energy company based out of UK fell victim to a wishing scam. What happens in that case is uh you get a phone call. So one of the employees uh in that company got a phone call and the phone call um said that I am the CEO and the voice actually looked like the CEO.

It was so specific. The CEO of that energy company uh was German. So the voice had a German accent and the person who who received the call could actually see that yes the voice has a German accent and sounded very familiar to the CEO himself. Um and he asked him to wire around $250,000 of money to uh um energy supplier in Hungary and he went ahead did that transaction and finally those that money was lost by the company. Unfortunately, in this case, uh the money didn't come back.

At least in the case of the casino, they were able to stop that attack, but not in this one. So, it brings me to the point that AI security in the age of AI is going to look like this where you are using the technology to defend ourselves, defend our infrastructures, defend our critical services. At the same time, attackers have also access to this really powerful technology that is giving them more power, giving them more resources to quickly uh understand their attacks, uh know more about their victims and conduct their cyber crimes. If you if you recall the the uh wishing example that I just uh said about uh all the attackers did was they performed social engineering on who they should be calling and they performed social engineering on who owns the company, what is the um you know uh demographics of the person who owns the company what what what are the interests and you know things like that that's how they were able to make that German connection and so on. So with by using a technology which is as powerful as AI, they were actually able to create an attack that looked very close to someone an like an actual human calling.

So um let's talk a little bit more about these two use cases now that we know that you know this this technology has uh both uh positives and negatives of it. I I kind of believe that cyber security is probably one field where the both the positive and negative aspects of artificial intelligence are very well documented. We have known cases of both the positive aspects of it as well as very well documented and known cases of negative aspects of it. Um so AI as our cyber ally that is one case where the industry is heavily uh utilizing the technology to improve our defenses. Uh one of the use cases is handling high volume of data.

Uh one example could be uh Google published a report saying that Gmail uh blocks 100 million uh uh fishing emails every day. So really high volume of data uh speedy threat detection quickly looking at threats analyzing them and and and blocking them in real time. Um but it also introduces new challenges challenges like def defakes and impersonation AI powered cyber crime. Just a week ago or two weeks ago um OpenAI published a report on how cyber crimes are cyber criminals are using uh technologies like chat GPT to better enhance their uh cyber crime operations. Uh there are also questions about adversarial attacks on AI that Max also just uh talked about a little bit.

Uh so what does it leave us to? It leaves us into an era where we are using AI to fight AI. Um it's it's it's going to be less about humans fighting with humans and it's going to be more about machines fighting with machines. So what is my closing thought over here? The closing thought is that our cyber defenses don't hinge on AI alone uh but on the wisdom and vigilance with which we use it.

It depends on how we are better using the technology for the good. But at the same time we know that these technologies are accessible to cyber criminal groups as well and we have to be very vigilant on making sure that we also understand how we better protect them. How we enforce better policies around it so that when these technologies are built up shared or distributed like the decentralized networks and so on there should be proper guardrails in place. That's all from my side. Thank you so much.

[Music]