Learning Library

← Back to Library

AI Prompted Malware & OT Patch Gap

Key Points

  • The podcast stresses that AI isn’t autonomously creating malware; rather, humans craft prompts that make AI generate more sophisticated code, so a sentient‑AI threat like HAL or Skynet is still far off.
  • New IBM Institute for Business Value benchmarks reveal a significant OT‑IT patching gap, with median high‑severity vulnerability remediation at about 90 % for IT but only 80 % for OT, and an even larger lag for medium‑severity issues.
  • Hackers are increasingly targeting industrial control systems in critical sectors (water, energy, agriculture), exploiting the IT‑OT convergence that exposes formerly isolated OT environments to internet‑based attacks.
  • A key obstacle to closing the OT patching gap is that many OT devices remain physically isolated; patching often requires hands‑on, on‑site intervention, which conflicts with the desire to keep these systems online and secure.
  • OT teams prioritize continuous uptime and safety over rapid patch deployment, leading to slower remediation cycles compared with the more agile IT patching processes.

Sections

Full Transcript

# AI Prompted Malware & OT Patch Gap **Source:** [https://www.youtube.com/watch?v=N_wGf6PAP5U](https://www.youtube.com/watch?v=N_wGf6PAP5U) **Duration:** 00:41:28 ## Summary - The podcast stresses that AI isn’t autonomously creating malware; rather, humans craft prompts that make AI generate more sophisticated code, so a sentient‑AI threat like HAL or Skynet is still far off. - New IBM Institute for Business Value benchmarks reveal a significant OT‑IT patching gap, with median high‑severity vulnerability remediation at about 90 % for IT but only 80 % for OT, and an even larger lag for medium‑severity issues. - Hackers are increasingly targeting industrial control systems in critical sectors (water, energy, agriculture), exploiting the IT‑OT convergence that exposes formerly isolated OT environments to internet‑based attacks. - A key obstacle to closing the OT patching gap is that many OT devices remain physically isolated; patching often requires hands‑on, on‑site intervention, which conflicts with the desire to keep these systems online and secure. - OT teams prioritize continuous uptime and safety over rapid patch deployment, leading to slower remediation cycles compared with the more agile IT patching processes. ## Sections - [00:00:00](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=0s) **AI Malware Prompt Debate & OT Patch Gap** - The episode debunks the myth of sentient AI creating malware, explores the AI‑assisted malware controversy, and highlights a significant lag in OT patching compared to IT. - [00:03:12](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=192s) **OT Security Challenges and Legacy Risks** - The speakers explain how outdated operational technology, a focus on uptime over comprehensive protection, and limited patching create predictable exposure to cyber and physical threats—highlighting the need for stronger security measures and robust contingency plans. - [00:06:44](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=404s) **Balancing OT Security and Risk** - The speakers debate whether operational technology must accept inherent security risks or can be hardened—through network segmentation, stricter patch governance, vendor accountability, and incident response—without hindering economic activity. - [00:10:39](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=639s) **Water Pressure Panic, Freight Hack** - The segment shifts from residents gathering on streets to discuss low water pressure to a report on a sophisticated cybercrime ring that hijacks freight‑marketplace communications to steal cargo. - [00:14:37](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=877s) **Shipping Wars of the Modern Age** - The speaker likens the surge in cargo and package theft—including hijacked trucks and potential OT system attacks—to a Wild West‑style episode of the show *Shipping Wars*, emphasizing rising incident rates and looming threats to major logistics firms. - [00:20:07](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=1207s) **Delayed Logic Bombs in npm Packages** - Researchers uncovered malicious npm packages that operate normally but contain hidden, time‑delayed logic bombs set to activate years later, initially causing random process terminations and then silently corrupting write functions. - [00:23:16](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=1396s) **Strategic Threat of Hidden Malicious Apps** - The team debates how covert malicious packages exploit user inattention, urging prioritization of this evolving, long‑term persistence threat despite uncertainty about its broader prevalence. - [00:26:33](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=1593s) **Questioning AI Ransomware Hype** - The speakers critique inflated claims about AI‑driven ransomware, note the retraction of a flawed MIT Sloan paper, and discuss how model safeguards versus self‑hosted bots shape the real threat landscape. - [00:31:26](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=1886s) **AI-Enhanced Malware and Media Hype** - The speaker explains that threat actors use AI prompts to refine existing malware—not to generate it autonomously—and questions whether cybersecurity reporting is overstating the danger for headlines and hype. - [00:35:00](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=2100s) **Weak Passwords in High‑Value Institutions** - The speakers highlight how major organizations, such as a museum housing billions in priceless items, often neglect basic security measures like multi‑factor authentication and strong password policies, a laxity that appears common among executives. - [00:38:59](https://www.youtube.com/watch?v=N_wGf6PAP5U&t=2339s) **Human Layer as First Defense** - The speakers stress that staff training, strict physical access controls, and basic cyber hygiene are essential lessons from a recent physical‑intrusion incident with cyber implications. ## Full Transcript
0:01It's not the AI writing the malware. It's actually more 0:04in the prompt the individual is using to get the 0:08AI to make the malware better. So I don't think 0:11it's sentient either. I don't think we're going to have 0:13a HAL 9000 situation or a Skynet situation come up 0:17anytime and probably my lifetime. Anyway. All that and more 0:21on Security Intelligence. Hello and welcome to Security Intelligence, IBM's 0:30weekly cybersecurity podcast where we break down the most interesting 0:34stories in the field with the help of our panel 0:36of experts. I'm your host, Matt Kaczynski, and joining me 0:39today, Claire Nunez, creative director, IBM X Force Cyber Range, 0:43Austin Zeisel, Threat Intelligence consultant and valiantly fighting an illness 0:48to be here, Dave Bales, X Force Incident Command and 0:51part of the not the Situation Room podcast. Thank you 0:53all for joining me. This week's stories, we're going to 0:56be talking about some hackers who are stealing real world 0:59cargo, a delayed malware, the AI malware versus AI slop 1:05debate, the Louvre's password, which I'm sure you've all heard 1:08about. But first, I want to get into the OT 1:12IT patching gap. Now, according to a new set of 1:20benchmarks from the IBM Institute for Business Value, OT patching 1:24rates lag pretty significantly behind IT patching rates. So, for 1:29example, the critical high vulnerability Patching rates in it are 1:32about 90% in the median for OT, it's 80%. So 1:35there's a 10 percentage point gap right there. When it 1:38comes to medium vulnerabilities, the gap's even bigger. You've got 1:4082% in it versus 70% in OT. And this comes 1:44at the same time that hackers are increasingly taking aim 1:47at industrial control systems in critical sectors like water, energy, 1:51and agriculture. These attacks can lead to some pretty serious 1:55safety issues. Right? We've seen things like hackers getting in 1:58and adjusting the chemicals used in water treatments that could 2:01potentially poison people. So these attacks are all enabled by 2:06the kind of IT OT merger, Right? The idea that 2:08more and more OT systems are now Internet enabled and 2:11they're part of the enterprise network, and that means they're 2:14exposed to these kinds of things. So I want to 2:17ask, you know, if we're bringing OT into it, how 2:20come our patching practices haven't caught up? Is there something 2:23wrong with our approach to this merger? And I want 2:25to throw it to you first, Dave. What do you 2:26think's going on here? OT typically tends to be offline. 2:31And I think that the problem with Patching OT is 2:34that someone actually has to get up from their desk 2:36and physically walk over and patch the OT systems. In 2:41a perfect world you wouldn't have to do that, but 2:43then they'd be online and they'd be even more susceptible 2:46to attacks from threat actors. So I think it's really 2:51just a matter of having to actually visit each OT 2:54system and patch it from there. That makes sense. Austin, 2:57any thoughts to add? Yeah, going off what Dave just 2:59said. So within OT environments they often prioritize that uptime 3:05and safety over rapid patching and speed, unlike IT systems. 3:12So many of those OT systems can simply just be 3:16rebooted or updated without really disrupting critical operations. And I 3:22think that operational mindset has created really a predictable window 3:27of exposure. Exposure that a lot of these adversaries have 3:31been exploiting. Absolutely. Claire, your thoughts? Yeah. So ot, as 3:35Dave mentioned, is a lot harder to go in and 3:39patch and make changes to. Another issue with a lot 3:42of the infrastructure, especially in the United States, is it 3:45is old. It's very old. We see that being. We 3:49see the impacts not just in terms of cybersecurity, but 3:52also weather. So a lot of our systems are very 3:55susceptible to weather. You see the issues from like the 3:59Texas snowstorms a couple years back. So you know, these 4:03systems are very delicate and we don't always, we, we 4:10prioritize availability over actually ensuring that they're secure in every 4:14sense. So physically secure, cyber secure, it's not, it's not 4:19easy to go in and take care of these operational 4:22systems. It's also very expensive. There are also OT platforms 4:27that kind of help with security, but a of organizations 4:29don't actually necessarily use them. So it is a little 4:34bit scary to think about too. And not to be 4:36kind of like a doomsdayer, but like everybody should have 4:40an alternate plan in the case that these systems, you 4:43know, go down for whatever reason. So. But it is 4:47really scary. And I think, Matt, the example you shared 4:49was lye being shared added to the water on accident. 4:54There's also been cases of pressure changing. So there's a 4:58lot of kind of ways that threat actors can manipulate 5:02the power grids and the water systems. And it's just 5:06something that not everybody is super cognizant of because they're 5:10not necessarily thinking about the security that goes into these 5:14different systems that we use every day. And a lot 5:19of these programmable logic controllers are 20, 30 years old 5:23too because like Claire and Austin both said it's not 5:28as easy as just patching these Things and having the 5:32uptime stay stable. Once you take these down, it takes 5:37a while for them to go down and then come 5:39back up. So patching them is probably lower on the 5:43priority scale than, say, getting the lye out of the 5:47water. And turning it off doesn't mean just turning off 5:51one element. You're likely turning off several different elements, and 5:54you have to patch one and you have to patch 5:56another, and then you have to make sure they're still 5:59connected to your IT and the IT OT connection is 6:02working properly. So it. It's not as easy as just 6:06turning something off to fix it and turning it back 6:08on. And this impacts all kinds of businesses, not just, 6:14you know, us on a human level for needing energy 6:18and water and all that kind of stuff. Like a 6:20bank cannot operate its data centers if it doesn't have 6:24any power. So I think a lot of other businesses 6:28aren't necessarily fully thinking about the OT impact on their 6:31own organizations as well. Yeah, I think those are some 6:34really good points that you folks bring up, especially around 6:36how it's a lot more complicated to do some of 6:39these patches, do some of these security fixes for OT 6:42than some of these, other than your average IT system. 6:44And Claire, you even said that, you know, you brought 6:46up the point that a lot of this stuff is 6:47also not having maybe the physical maintenance that IT needs 6:50or those kinds of security issues. And so that brings 6:53me to maybe this is a naive question for, you 6:56know, because I'm not somebody who's in this realm, but, 6:58you know, is it just the case then, that we 7:00have to accept this level of risk, that this is 7:02just kind of. It comes with ot? This is just 7:04what it's like. Like the patching will always be less 7:07than IT or. I don't know, could we ever get 7:09to a point, you know, could we change things where 7:12we can actually maybe make these systems a little more 7:14secure, get those investments in without necessarily harming economic activity? 7:18Which I understand is a huge question, but I'll ask 7:20Austin. Austin, do you have any thoughts? Are we just 7:22kind of doomed to accept this level of risk? It 7:24comes down to a matter of prioritizing those security measures, 7:28and that first begins with network segmentation and passive monitoring. 7:34Those are really critical, especially in OT environments, but they're 7:37really only half the story. I think leaders in that 7:41space must also pair those controls with stronger patch governance, 7:46because we just talked about earlier how important that patching 7:49is also accountability with vendors, because with ICS systems, ot, 7:54you have so many different third parties and supply Chain 7:57vendors involved. So true. OT specific incident response planning is 8:04really critical here. Not just repurposing it from, you know, 8:09outdated IT playbooks. I like that a lot. You need 8:12a very specific approach to this kind of thing. You 8:14know, we might be making OT and it look similar 8:16in a lot of ways, but they're still very different 8:18systems and you need a very different approach. And Claire, 8:21I just wanted to ask you, another point that you 8:23had brought up was about how these OT systems are 8:25often entangled in so many other systems. And that's part 8:28of what makes it so complicated when it comes to 8:29patching them. You're not just shutting down like one machine. 8:32You can shut down like a whole, you know, production 8:35facility basically to do this. I'm wondering if you have 8:37any thoughts on how that feeds into our ability to 8:41maybe someday get this, you know, problem under control. I 8:44think it plays into it if we're only thinking about 8:48today. And I don't think that's the best way of 8:51looking at resiliency for these systems. Especially, again, we are 8:56looking at multiple challenges in terms of cybersecurity, physical security 9:01from natural disasters, and people sneaking into these locations and 9:05such. And I think, you know, the natural reality is 9:08that most organizations are just thinking about, you know, the 9:14near term in terms of some of these systems because 9:16the investment is so large. But the investment will be 9:20larger if you have to take your plants or your 9:23grids down for a very long time and you have 9:26constituents who are angry, citizens who are angry, or you 9:30can't produce at all. For some organizations, the cost of 9:35not being able to produce is in the millions per 9:38day. So it's kind of thinking, okay, if we have 9:42to take down our site for 24 hours, that's going 9:45to cost us this much. Plus the cost of changing 9:49everything within there versus being down for potentially two weeks, 9:53three weeks because of some kind of damage, and then 10:04backups or something, or like a natural disaster. So it's 10:08doing some cost benefit analysis. And I don't know if 10:11everybody's necessarily doing it because it's very expensive no matter 10:15what. It's just one much larger price point than the 10:21other. How would everyone feel if they just, for just 10:24the smallest example, lost water pressure for a couple of 10:30hours? People are going to complain about that. Yeah, people 10:32will absolutely. Just water pressure. Forget the lye, forget the 10:36fluoride, forget turning the water off, just lose water pressure. 10:39For a couple of hours and listen to your neighbors 10:42all start coming out into the middle of the street 10:43and talking about the fact that there's no water pressure. 10:46I was going to say, Dave, literally like a year 10:48ago, that's it. That happened in my neighborhood. And you're 10:50right, it's just water pressure. We're not even talking about 10:52something that's like life or death. It's just everyone's suddenly 10:55out in the streets. Is your water pressure working? Is 10:57your water working? Is it just my water? And yeah, 10:59so, you know, that does. You're right. That, that, that 11:01complicates things even more. You know, I guess, I guess 11:04we'll round out the segment by saying that I don't. 11:08I feel bad for the folks who have to make 11:09these decisions because Lord knows I couldn't make them. Let's 11:12move on to the next story. Hackers steal real. Now, 11:22speaking of cyber attacks that have some very material consequences, 11:26proofpoint reported last week on a sophisticated cybercrime ring that's 11:31targeting freight companies to steal their cargo. Now a real 11:34quick summary of how freight companies connect with carriers might 11:38be in order for those who aren't familiar with it, 11:40they tend to use a kind of marketplace system, right? 11:42So a freight company comes in and says, hey, we 11:44have this load. Carriers come and say, hey, we can 11:46carry it for this amount. They make a deal. It 11:48works that way. The way this cyber attack operates is 11:51that first the hackers get in there and they impersonate 11:54a freight company and say they have a fake load, 11:57carrier comes and bids on it. They use that opportunity 11:59to compromise the carrier's account, and then they act as 12:02the carrier and they bid on real loads and say, 12:05hey, we can take this job. When they secure the 12:08job, they send their own trucks over to pick up 12:10the load and take it. And of course, it never 12:13makes it to its final destination. I just, the brazen 12:16ness of this really struck me, which is why I 12:18wanted to kind of bring it up and see how 12:20folks feel about it. Claire, let me start with you. 12:22Any initial reactions to this cyber attack? I kind of 12:25think about it and in terms of like, that's such 12:29a mixed bag where you're just stealing people's Amazon packages. 12:33Like you can either get a computer or you're getting 12:35just soap or something. Where, I mean, I guess with 12:40cargo, it's, I'm, I'm like, initially I was like, oh, 12:43that's interesting. Like, how do you get rid of all 12:46of that cargo that's within? But I thought it was 12:49just an interesting overall concept where it's just like you're 12:54reselling these items and it's. You have to be really 12:59organized to take on that kind of level of crime 13:04and to kind of figure out what you're going to 13:07do with all those shipments as well. When I was 13:09reading kind of the Reddit thread that you sent over, 13:12Matt, about the person whose organization had faced this, it 13:16came down to these people receiving an email saying, you 13:22have to do this for a specific order and then 13:24they input their information. And that's kind of how the 13:27threat actor got in. And it's like, of course, in 13:30my mind, it came down to somebody rushing and clicking 13:33a link they shouldn't have. And I wonder how much 13:36more common this is than reported, because shipping is now 13:41like, everything has to be shipped within two days kind 13:44of thing. So there's probably just a tremendous sense of 13:49pressure on these people that, like, sending them a phishing 13:54email like, that would definitely get them to act. So 13:58I wasn't surprised seeing that. But it's interesting to see 14:02because sometimes I'm guessing, like, people will get packages delayed 14:06and such, and you don't necessarily know about why. And 14:10maybe it's because it was actually just lost by somebody 14:14in a warehouse or it was part of a larger 14:18cargo cyber crime ring. Yeah, it is funny, right? It 14:21is kind of like a scaled up version of the 14:23mail theft ring. I didn't really think about that, but 14:25it's, it's extremely obvious. And also, side note, one of 14:28my proudest moments was foiling some male thieves in my 14:30neighborhood. But we don't have to talk about that right 14:32now. Good for you. You were going to, you were 14:35going to say something. Come on in here. I was. 14:37I feel like I'm watching a true life episode of 14:40Shipping Wars. I don't know if anyone's ever seen the 14:42show. Yes, I have. But that's what this feels like. 14:46It feels like Shipping wars, except it's the old west 14:49version of Shipping wars where they're jumping on trains and 14:53stealing things. But no, like Claire was saying, it's, you 14:56know, when you start talking about whether or not your 14:59package has been lost or whether it's been stolen. I've 15:03actually had packages stolen from the truck, not from my 15:07doorstep. It will say, oh, well, your package has been 15:11delayed, and we'll let you know. If it doesn't arrive 15:13within two days, let us know. Well, so these pirates, 15:21these thieves, these people that do this, I don't understand 15:26what the end goal is. Aside from making money and 15:30making other people miserable. And they do. They make people 15:33miserable. That could be a load of diapers for a 15:38new grandfather, let's say. Or it could be, like Claire 15:41said, a computer. Why are they doing this? Because it's 15:45easy to do. That's why. The attack numbers have increased 15:48year over year. Like 2024, I think it was 25, 15:5327%. And then projected for this year, it's another 22% 15:57increase. I mean, it's. They're going after large transport firms 16:01and it's just a matter of time before they hit 16:02an Amazon truck and then we're all in trouble. Now 16:05I'm sort of wondering too, in light of that last 16:08story that we were talking about, you know, in terms 16:10of the OT systems being attacked, and now we're talking 16:12about these cargo theft getting into the mix. It almost 16:15feels to me like our concept of cyber attack needs 16:18a little bit of updating. And again, maybe this is 16:21just myself and my own projection, but I think cyber 16:23attack and that cyber part comes in pretty big, right? 16:26You still kind of think this stuff is happening on 16:28computers, but it's spilling out into the real world in 16:30these very interesting ways for me. And I just, I'm 16:32wondering if, I don't know, do other people kind of 16:36feel the same thing going on? Or maybe you've been 16:37thinking about this much longer than I have. Austin, let 16:41me, let me throw it to you. Have any thoughts 16:42on this kind of blurring of the lines between real 16:45world security, physical security and cybersecurity? Yeah, that's an interesting 16:48point because, you know, the tides have shifted. You know, 16:52we think of cyber attacks as strictly being digital and 16:56data oriented, whereas these, these attacks are having real world 17:02physical implications. And I don't think people, at least the 17:06general public, fully fully understand that, especially when there's so 17:10many different third parties involved, smaller brokers, multiple carriers within 17:16those supply chains. So that creates much more fragility amongst 17:22those supply chains in this logistics and transportation sector. Absolutely. 17:27Claire, any thoughts on your end? Yeah. In this cyber 17:30range, we've increasingly seen clients go from being more focused 17:34on the loss of data, which is an awful thing 17:36for a lot of US consumers, to more like material 17:40impacts of having either impacting shareholders, employees, and again, customers 17:45in terms of service availability. So I think it's also 17:48a bigger concern for organizations too, where it's like, I 17:52will not be able to produce X products and there 17:56will be an outage for a certain amount of time 17:58and then all of our customers are going to switch 18:00to Y product because it's not available. So I think 18:05as a consumer we also don't care as much about 18:08our data being lost because our data has been lost 18:11how many times. I'm sure everybody on this call has 18:13like, you know, at least 10 years of data credit 18:17monitoring. So you know, like, I think it's something that 18:23organizations are also thinking about because they're scared because that's 18:26more of a, like an impact to their business if 18:29they can't produce or provide services and that's going to 18:33impact them more. And I think at the end of 18:35the day, if you can't kind of going back to 18:37the OT, if you have to pay $2 million to 18:41get your site back up and running and it costs 18:445 million a day of downtime, you're going to pay 18:47the 2 million in ransom. You might not admit it, 18:50but you're, you're going to pay it because it's, it's 18:53again the cost benefit analysis. So you're either going to 18:56lose X number of customers and it's going to cost 18:59you this much or you know, so I think it's 19:03just kind of a, it's a mental shift a lot 19:05for businesses and then consumers were starting to get annoyed 19:08by it as well as it trickles down. I'm going 19:11to age my 50 year old self here. Back when 19:14I got my first computer back in the early 90s, 19:18we'll go with that. No one ever thought about the 19:22number of IP addresses that we had or had available 19:25or whether or not we were going to run out 19:27as our landscape shifts to having computers control everything. And 19:33let's be honest, it does. Everything is a cyber attack 19:37now. So we can't just think of cyber attack as 19:41one person attacking a computer. You have to put that 19:44the supply chain, that's a cyber attack. You know, the 19:47physical attacks, yeah, those are different but for the most 19:51part we're looking at cyber attacks on everything. So yeah, 19:54our definition definitely needs to be updated for cyber attacks 19:58right now. It's still the old school definition and we 20:01need to get away from that and towards a more 20:04modern definition. Move on to our next story here folks. 20:08This is malicious packages that drop time delayed logic bombs. 20:17Researchers at Socket identified nine malicious nougat packages with hidden 20:22logic bombs designed to detonate years after they've been installed 20:26and working properly the whole time. So these packages, they 20:29do work, they do what they say on the tin, 20:30but hidden inside of them are these logic bombs that 20:33are set to detonate, go off after 2027 or 2028, 20:38depending on what package you're dealing with. And the way 20:41they work, kind of their malicious components are that, you 20:45know, when you first install them, they will start immediately 20:47terminating some processes. But they only do it with a 20:5020% chance because it's designed to kind of make it 20:52seem random, like it's not related to the package, it's 20:54just something crashed. But then the specific kind of code 20:58that detonates after a few years is they start silently 21:02corrupting write functions. Right. And so I just. The thing 21:06for me was trying to figure out what's the point 21:08of the time delay, what does the attacker gain, what's 21:11the play here? And I want to throw to you, 21:13Austin, what do you think about this technique? Well, first 21:15off, I've actually never even heard of this technique being 21:19used. So it's pretty, pretty wild that this is going 21:23on. But as you're asking, what's the point here? Well, 21:26I think it's really about the dependency on time that 21:31because this gives attackers so much dwell time, by the 21:34time say the payload is triggered, that initial dependency has 21:39been forgotten because so much time has passed already. So 21:44attackers really aren't thinking and days, weeks or months, they're 21:47thinking in years. And that's where they're using time to 21:52their advantage and these threats are becoming successful. It's funny, 21:55right? The attackers have the long term thinking, but we 21:58talked quite a bit about how our organizations don't always 22:01have the long term thinking. Right. When it comes to 22:04OT patching. There's a little bit of an asymmetry here 22:07that is, I don't know, something to think about. Dave, 22:10any thoughts on your end about this technique? Yeah, so 22:13what they're looking for is complacency. How many times have 22:16you installed a piece of software and five years later 22:20it's still sitting there? Whether you use it or not 22:22is irrelevant, but it's still sitting there. I personally probably 22:26don't uninstall about 90% of the software that I install. 22:30So if I'd have downloaded one of these NuGet packages 22:34in couple of years, I'm going to have a logic 22:36bomb go off on my machine because I just don't 22:40do it. I don't uninstall things. And I think that's 22:42what these attackers are counting on. They're going to get 22:45these packages installed. They've been downloaded what, 9,500 times. So 22:519,500 people who haven't read this story are going to 22:54be in for a very big surprise come September of 22:5827 and October of 28th. That's a really good point. 23:02I mean, I'm looking at my own desktop right now. 23:03It is littered with things I've downloaded I haven't touched 23:07in forever. And you're right, I don't bother to get 23:09rid of them. Storage isn't really a problem anymore. So 23:12you download a thing and you just let it sit 23:14there. That's a good point. I hadn't thought about that. 23:16I should delete some apps. Claire, what about you? Any 23:18thoughts on your end? I just think it's smart from 23:21a threat actor perspective. It's just. It's smart. Most people 23:25will forget about things. Most people won't even read that 23:28headline and know that there's something they shouldn't have downloaded 23:33in that. In whatever they downloaded. So I just think 23:36it's really like a innovative way of getting malicious packages 23:43on people's devices without them knowing, even though they, at 23:47this point, I guess, should know since it's out there. 23:50But it's just smart from a threat actor perspective. Yeah. 23:54Austin, I want to come back to you again because 23:55as you had said, right, this is something you hadn't 23:58really seen before. This is kind of a new one 23:59for you. I'm wondering, do we need to be paranoid 24:03about this kind of thing lurking in more packages? Do 24:06we need to be worried about this or do you 24:08think this is one incident? How should we approach this? 24:11How should we think about this is kind of threat 24:13on a scale level? Any thoughts there? Yeah, I mean, 24:15I think we should prioritize this threat because again, the 24:19threat landscape's always evolving and we're seeing this play out. 24:23And so we need to sort of think ahead of 24:26the threat actor because there is some strategic intent here. 24:30This is about long term persistence and also leverage by 24:33using time to their advantage. And again, it's not just 24:37about stealing data. It's about planting those seeds for future 24:41leverage and control as well as just overall sabotage. Gotcha. 24:45Dave, any thoughts on your end about how defenders should 24:48be thinking about this kind of thing? Typically, I would 24:51think that the defenders of databases and industrial controls would 25:04wipe it out completely. I mean, now you've got four 25:08years worth of data that you can play with. Industrial 25:12controls, same thing, you've got four years worth of data 25:16to play with. So you're in for some really big 25:19disruptions if we don't get all of these traced or 25:23as many as we can traced to the downloaders and 25:27get them off of those systems because they're probably going 25:30to extremely important functions. Absolutely. No, that's a very good 25:35point and I neglected to mention, but yeah, these packages 25:38target largely database systems and icss, and so yeah, they 25:42can cause some serious problems there. Let's move on then 25:45to the next topic. And this is honestly maybe the 25:48one I'm most excited about to ask you folks about 26:04reports of AI powered malware that we've seen come out. 26:06And in the last week, two big instances happened that, 26:10that made me want to bring this before you folks 26:11today. The first was that Google started talking about an 26:14experimental malware it found called Prompt Flux, which can interact 26:18with Gemini's API to regularly request rewrites to its code 26:22with the intent of evading detection. But as cybersecurity researcher 26:26Marcus Hutchins points out, this self modification function was commented 26:31out in the code, so it wasn't even actually active. 26:33And even if it were, it probably wouldn't operate all 26:36that seamlessly. Just because it's requesting, you know, code rewrites 26:39doesn't mean those rewrites are actually going to evade any 26:41detection system or they might not even work. The other 26:44thing that happened was that there was a widely cited 26:46MIT Sloan working paper that came out in the spring 26:49and it claimed that 80% of active ransomware attacks were 26:53using AI. It has since been withdrawn after some serious 26:57criticism came out regarding the methodology, with researchers wondering how 27:01the authors were able to definitively say AI was involved 27:04in any of these attacks. So my question is, have 27:08we lost the plot when it comes to these AI 27:11powered threats? You know, are we getting a little ahead 27:13of ourselves with the way we talk about these things? 27:16Dave, what are your thoughts here? So there are built 27:18in protections in, in all the AI models for the 27:22public iterations of these, these AI models that keep you 27:26from putting your code in and getting some more malicious 27:30code out. However, you can take the training wheels off 27:33of these and put your own instance in. And now 27:37you've got your own little personal AI bot who is 27:41willing to do whatever you want. And as long as 27:45we, and we talked about this on the other show, 27:49once you start removing the gates, you start allowing more 27:54things to happen, we're going to see this more and 27:56more, the research paper from MIT notwithstanding. I mean, they 28:00did make some grandiose claims, but you are going to 28:05see AI malware come into existence and it's going to 28:09be dangerous and it's going to be good and it's 28:11going to be a way for adversaries to get into 28:16systems that they wouldn't normally have the skill set to 28:19get into. But, but AI absolutely has the skill set 28:23to do whatever it needs to do, programmed by humans. 28:28It's a lovely partnership. You know Austin, your thoughts on 28:33this kind of subject? Do you think we are maybe 28:35a little sensational about this or do you think this 28:37kind of thing is something to be worried about right 28:39now? I mean, I think AI, the idea of AI 28:43being fully autonomous, self evolving malware is sort of in 28:47the realm of science fiction than reality. Because at the 28:51end of the day, AI is a tool created by 28:54humans and it's really a reflection of us. It's not 28:57independent of, of human nature. Perhaps that day will come. 29:03But I think for now that isn't fully operational. You 29:08know, I don't think AI is going to become conscious 29:11and self aware. So I don't really see a 2001 29:14Space Odyssey scenario playing out. But maybe I'm just being 29:18too optimistic. So. Yeah, yeah. And it's kind of like 29:23Dave said, right? The threat necessarily isn't the sentient malware, 29:27but it's like if a hacker has their own little, 29:29you know, training wheels off model to help them do 29:33things they couldn't do otherwise. Right. It's, it's sort of 29:35less sensational, but just as dangerous, maybe more dangerous. Claire, 29:40your thoughts on this kind of situation? Yeah, I agree 29:43with Austin. I don't think it's something, it is very 29:47sci fi right now. But I do think it would 29:50be a mistake to say that threat actors aren't using 29:53AI in some capacity. I mean, again I mentioned in 29:56like, I think our last episode that I was on, 30:05So they are also trying to optimize their operations, which, 30:08which probably includes bringing in AI to that. So I 30:12mean, it would be kind of silly for a threat 30:15actor to not be using AI to check their code 30:19or to write phishing emails or something. It would just 30:23extremely speed up their operations. So I think it's safe 30:28to assume that most of them are using AI in 30:30some capacity like that. But I don't know if we 30:32fully have gotten to the point of like self evolving 30:37malware that is, is almost like a real like biological 30:43virus of some kind. I. We're just, I don't think 30:46we're there yet. And if we are I don't, I 30:51don't know if we'd pick up on it yet. But 30:53I, I mean, I don't necessarily think it's like something 30:58super true yet, but I think a lot of people 31:01are like, forget that AI can be a little scary 31:06if you are trying to do prompt injections and trying 31:11to evade the safeguards. I agree that I don't think 31:14it's sentient yet. And I don't think that anyone would 31:18be foolish enough to think that AI is writing malware 31:22itself. It's all human based. I think it's more code. 31:28What is the word code embellishment that AI is being 31:31used for looking at the code that, that they're putting 31:35in. Hey, make this better for me. So it's not 31:37the AI writing the malware. It's actually more in the 31:41prompt that the individual is using to get the AI 31:46to make the malware better. So I don't think it's 31:49sentient either. I don't think we're going to have a 31:52HAL 9000 situation or a Skynet situation come up anytime 31:56and probably my lifetime anyway. But they are using it 32:01to make malware better because they do have these sandbox 32:05AI instances that allow them to do things that they 32:10wouldn't normally be able to do on say, Gemini or 32:13Copilot or chatgpt or anything else like that. Yeah. And 32:17I want to ask about something else that I've seen 32:19cybersecurity researchers increasingly talk about, is that some of this 32:23might be a kind of communication issue. And by that 32:26they mean some of the organizations reporting on this malware. 32:29They might feel a little, I don't know, motivated to 32:31make it seem more impressive than it actually is for, 32:34you know, first of all to get some headlines, to 32:36get some clicks. But also because if you're helping to 32:39kind of secure AI systems, you might have an investment 32:41in making the AI systems look super, super spooky. So 32:45I'm kind of wondering if you folks feel like we 32:48might be dealing with partially a communication issue here. Are 32:51we just like sensationalizing things too much? Is that. What 32:54do you think is going on? And I'll start with 32:56you, Austin. What do you think about that? This recent 32:58report that was what redacted it has definitely been sensationalized. 33:03Because if a board believes, you know, AI is driving 33:0780% of, of ransomware, which the stat they put on 33:11the headline, you know, they may start overspending and misallocating 33:15resources on AI detection tools, which will also underspend on 33:21basic security hygiene like Patching, segmentation and identity and access 33:27management. So I think that misallocation of resources can become 33:32a massive strategic risk for organizations. Cool. All right, well, 33:37then we will wrap that section up and move on 33:39to our final story of the week. Oh, my favorite. 33:43The password for the Louvre's video surveillance system was Louvre. 33:52By now, we've all heard about the theft of some 33:55$100 million worth of jewels from the Louvre. And in 33:58the wake of this incident, some news outlets have resurfaced 34:01some old reports on the Louvre's security measures, including warnings 34:05from the National Cybersecurity Agency about its kind of weak 34:08systems and the fact that at one time, at least, 34:10the museum's surveillance system password was Louvre. Now, Dave, given 34:14that you were so eager to dive into this one, 34:16I feel like I should just throw it to you 34:18first. Give me your thoughts here. What do you think? 34:20What do you think were they thinking? Honestly, password 1, 34:272, 3 would have been better, you know, and I'm 34:31sure they probably tried that and said, wait a minute, 34:33let's just go with the most simple thing and let's 34:35just see if it's luv. And they got in with 34:37that. And this was the camera system, right? This wasn't 34:42like the password to the Louvre cyber security system. It 34:47wasn't. I'm sorry, their security system, it wasn't keeping the 34:51jewels safe. It's just the camera system. Just the camera 34:55system? Really? I mean, that's a pretty big system there. 35:00I mean, this is a building that houses hundreds of 35:04billions of dollars worth of priceless. I mean, you can't 35:08put a price on it of priceless items. And they 35:12couldn't be bothered to put multi factor authentication in there. 35:16They couldn't be bothered to come up with something was 35:18more than six characters long. They couldn't be bothered to 35:21put even an exclamation mark at the end of it. 35:25Really? Or. You say for the E speak in there? 35:31Yeah, exactly. We need leet speak in there so that 35:34at least someone with half a brain would have to 35:37use it. Nope. Nope. My son could have probably done 35:40this and he doesn't know anything about computers. Claire, I 35:43wanted to ask you as somebody who, you know, I 35:46know you spend a lot of time working with organizations, 35:48training. I'm sure the Louvre is not the only institution 35:52that's been guilty of these kinds of lax sort of 35:54security processes. And I'm not asking you to tell us 35:57about other ones you've seen, but I'm wondering in general, 35:59have you noticed these kinds of lax Measures in places 36:02before. Is this a trend? I think it's pretty common, 36:05but a lot of people will not admit to it. 36:08They won't admit that their password is admin or password 36:12or whatever the username is, but I think it's very 36:17common. But a lot of the people that we have 36:20that come through the range probably wouldn't admit that. And 36:25these are, you know, executives, so they're not always like 36:28the most tech savvy folks either. They pro. They probably 36:31will not admit that that's their password as well. This 36:37is just, I feel like the password and the surveillance 36:40camera thing, that's like one of the first things you, 36:42you do when in a heist movie is you disable 36:47the camera. So it's like, even if the security association 36:51or the national whatever coalition it was didn't say you 36:55should do something about this, if you watch any of 36:58those kind of movies, wouldn't, you know. But it also, 37:02to me, this whole thing makes the heist so much 37:05less impressive because, I mean, just seeing all of the, 37:10like, lax security measures they had, it's like I. Someone 37:15could have a lot. Anyone could have done it. And 37:17I don't want to say anyone could have done it 37:18because it's definitely something that's very difficult. But if you 37:21look at like the Isabella Stewart, Stewart Garden, like heist 37:24in the 90s in, in Boston, that was just like, 37:28so well thought out. They still don't know who did 37:31it. But it's like with this one, they just kind 37:35of went in and they like typed in Louvre and 37:37called it a day. And they like, they used their 37:40drill and they took the jewels and walked, walked away. 37:43So. And we didn't find them yet. So it's as 37:46far as I know, I think there's suspects in. But 37:49still, I believe they're still at large. But yeah, they 37:51just kind of wore construction uniforms and walked in and 37:55nobody really stopped them until they were taking jewels and 37:57at which point it was over, you know. Austin, any 38:00thoughts on your end? This was a matter of not 38:04being a technical flaw, but it was a flaw in 38:07human behavior and governance. I mean, that's really the big 38:11failure there. Sadly, convenience typically overrides policy when it comes 38:17to cybersecurity. And without any accountability, those critical systems remained 38:24pretty defenseless, especially with CCTV cameras. So kind of crazy 38:31that, you know, an institution as big as the Louvre, 38:35world renowned, thought maybe they were too good for any 38:39password hygiene or password rotation. Also, like, how many people 38:44pass through the Louvre a day? I'm sure. There are 38:47other devices that are just available that you could type 38:50that password into and get in. Like, how many people 38:54are in there that. That, you know, could. Something's not 38:56watched and you could just type it in. It's scary. 38:59You know, I want to end on a constructive note 39:01here. So I'm wondering, you look at a situation like 39:04this, what kind of lessons you think we can learn 39:06from. From the lose, you know, maybe lax approach to 39:10some of their cyber security measures. Let's start with you, 39:14Claire. Any thoughts there? There's just so many things that 39:17they could have. Could have done. I feel like training 39:20of their staff as well would have helped too. But 39:26I. I think this whole thing goes back to when 39:28we earlier said, or Dave earlier said, like, everything is 39:31cyber now. So, like, this wasn't a cyber attack. It 39:33was a physical attack, but there is cyber involved. So 39:36it's just like, I don't know, have your folks trained, 39:39have better passwords. I, you know, don't let people just 39:44drive up to your institution with, like, a van with 39:47a ladder on it and go through the window. I 39:49mean, it's just. Claire has ruined every heist movie I've 39:55ever seen now. Ocean's Eleven will never be the same. 39:59I'm sorry. True. I was just gonna say, you wear 40:03a shiny vest and carry a clipboard into anywhere and 40:06no one bats an eye. Right. So it really comes 40:09down to the human layer as. As the first line 40:11of defense there. I like that. I like that it 40:13comes out of the human layer. I also just want 40:15to say real quick, I do think not letting people 40:16drive up to your. Anything with a ladder is a 40:19good, you know, thing to do in general. Like you 40:21said, Claire, really great advice to Dave. Any thoughts on 40:25your end in terms of lessons we can learn here? 40:27The typical answer is practice good password hygiene. I can 40:31almost assure you that whomever installed the cameras at the 40:37Louvre put the Louvre in as password, and no one 40:40ever thought to go back and change it, least of 40:43all the people who were in charge of it. They 40:45just thought, we'll get to that. And getting to that 40:49was the next day, and then it was the day 40:51after, and then it was the day after, and I 40:52think it just became forgotten. I don't think that the 40:57Louvre purposely used the easiest password that they could find. 41:02I think it was just forgotten. And I think they 41:07probably learned their lesson and it cost them a few 41:09bucks. I think that's a really good place to end 41:12it on. And it is all the time we have 41:14for today, folks. So I want to thank you, Dave, 41:16Austin, Claire, for sitting here and having this conversation with 41:19me. Thank you to the viewers and the listeners. As 41:21always, folks, subscribe to security intelligence wherever podcasts are found. 41:25Stay safe out there. And I don't know, just change 41:27your passwords up, man.