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
- 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.
- OT Security Challenges and Legacy Risks - The speakers explain how outdated operational technology, a focus on uptime over comprehensive protection, and limited patching create predictable exposure to cyber and physical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.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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
# 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
It's not the AI writing the malware. It's actually more
in the prompt the individual is using to get the
AI to make the malware better. So I don't think
it's sentient either. I don't think we're going to have
a HAL 9000 situation or a Skynet situation come up
anytime and probably my lifetime. Anyway. All that and more
on Security Intelligence. Hello and welcome to Security Intelligence, IBM's
weekly cybersecurity podcast where we break down the most interesting
stories in the field with the help of our panel
of experts. I'm your host, Matt Kaczynski, and joining me
today, Claire Nunez, creative director, IBM X Force Cyber Range,
Austin Zeisel, Threat Intelligence consultant and valiantly fighting an illness
to be here, Dave Bales, X Force Incident Command and
part of the not the Situation Room podcast. Thank you
all for joining me. This week's stories, we're going to
be talking about some hackers who are stealing real world
cargo, a delayed malware, the AI malware versus AI slop
debate, the Louvre's password, which I'm sure you've all heard
about. But first, I want to get into the OT
IT patching gap. Now, according to a new set of
benchmarks from the IBM Institute for Business Value, OT patching
rates lag pretty significantly behind IT patching rates. So, for
example, the critical high vulnerability Patching rates in it are
about 90% in the median for OT, it's 80%. So
there's a 10 percentage point gap right there. When it
comes to medium vulnerabilities, the gap's even bigger. You've got
82% in it versus 70% in OT. And this comes
at the same time that hackers are increasingly taking aim
at industrial control systems in critical sectors like water, energy,
and agriculture. These attacks can lead to some pretty serious
safety issues. Right? We've seen things like hackers getting in
and adjusting the chemicals used in water treatments that could
potentially poison people. So these attacks are all enabled by
the kind of IT OT merger, Right? The idea that
more and more OT systems are now Internet enabled and
they're part of the enterprise network, and that means they're
exposed to these kinds of things. So I want to
ask, you know, if we're bringing OT into it, how
come our patching practices haven't caught up? Is there something
wrong with our approach to this merger? And I want
to throw it to you first, Dave. What do you
think's going on here? OT typically tends to be offline.
And I think that the problem with Patching OT is
that someone actually has to get up from their desk
and physically walk over and patch the OT systems. In
a perfect world you wouldn't have to do that, but
then they'd be online and they'd be even more susceptible
to attacks from threat actors. So I think it's really
just a matter of having to actually visit each OT
system and patch it from there. That makes sense. Austin,
any thoughts to add? Yeah, going off what Dave just
said. So within OT environments they often prioritize that uptime
and safety over rapid patching and speed, unlike IT systems.
So many of those OT systems can simply just be
rebooted or updated without really disrupting critical operations. And I
think that operational mindset has created really a predictable window
of exposure. Exposure that a lot of these adversaries have
been exploiting. Absolutely. Claire, your thoughts? Yeah. So ot, as
Dave mentioned, is a lot harder to go in and
patch and make changes to. Another issue with a lot
of the infrastructure, especially in the United States, is it
is old. It's very old. We see that being. We
see the impacts not just in terms of cybersecurity, but
also weather. So a lot of our systems are very
susceptible to weather. You see the issues from like the
Texas snowstorms a couple years back. So you know, these
systems are very delicate and we don't always, we, we
prioritize availability over actually ensuring that they're secure in every
sense. So physically secure, cyber secure, it's not, it's not
easy to go in and take care of these operational
systems. It's also very expensive. There are also OT platforms
that kind of help with security, but a of organizations
don't actually necessarily use them. So it is a little
bit scary to think about too. And not to be
kind of like a doomsdayer, but like everybody should have
an alternate plan in the case that these systems, you
know, go down for whatever reason. So. But it is
really scary. And I think, Matt, the example you shared
was lye being shared added to the water on accident.
There's also been cases of pressure changing. So there's a
lot of kind of ways that threat actors can manipulate
the power grids and the water systems. And it's just
something that not everybody is super cognizant of because they're
not necessarily thinking about the security that goes into these
different systems that we use every day. And a lot
of these programmable logic controllers are 20, 30 years old
too because like Claire and Austin both said it's not
as easy as just patching these Things and having the
uptime stay stable. Once you take these down, it takes
a while for them to go down and then come
back up. So patching them is probably lower on the
priority scale than, say, getting the lye out of the
water. And turning it off doesn't mean just turning off
one element. You're likely turning off several different elements, and
you have to patch one and you have to patch
another, and then you have to make sure they're still
connected to your IT and the IT OT connection is
working properly. So it. It's not as easy as just
turning something off to fix it and turning it back
on. And this impacts all kinds of businesses, not just,
you know, us on a human level for needing energy
and water and all that kind of stuff. Like a
bank cannot operate its data centers if it doesn't have
any power. So I think a lot of other businesses
aren't necessarily fully thinking about the OT impact on their
own organizations as well. Yeah, I think those are some
really good points that you folks bring up, especially around
how it's a lot more complicated to do some of
these patches, do some of these security fixes for OT
than some of these, other than your average IT system.
And Claire, you even said that, you know, you brought
up the point that a lot of this stuff is
also not having maybe the physical maintenance that IT needs
or those kinds of security issues. And so that brings
me to maybe this is a naive question for, you
know, because I'm not somebody who's in this realm, but,
you know, is it just the case then, that we
have to accept this level of risk, that this is
just kind of. It comes with ot? This is just
what it's like. Like the patching will always be less
than IT or. I don't know, could we ever get
to a point, you know, could we change things where
we can actually maybe make these systems a little more
secure, get those investments in without necessarily harming economic activity?
Which I understand is a huge question, but I'll ask
Austin. Austin, do you have any thoughts? Are we just
kind of doomed to accept this level of risk? It
comes down to a matter of prioritizing those security measures,
and that first begins with network segmentation and passive monitoring.
Those are really critical, especially in OT environments, but they're
really only half the story. I think leaders in that
space must also pair those controls with stronger patch governance,
because we just talked about earlier how important that patching
is also accountability with vendors, because with ICS systems, ot,
you have so many different third parties and supply Chain
vendors involved. So true. OT specific incident response planning is
really critical here. Not just repurposing it from, you know,
outdated IT playbooks. I like that a lot. You need
a very specific approach to this kind of thing. You
know, we might be making OT and it look similar
in a lot of ways, but they're still very different
systems and you need a very different approach. And Claire,
I just wanted to ask you, another point that you
had brought up was about how these OT systems are
often entangled in so many other systems. And that's part
of what makes it so complicated when it comes to
patching them. You're not just shutting down like one machine.
You can shut down like a whole, you know, production
facility basically to do this. I'm wondering if you have
any thoughts on how that feeds into our ability to
maybe someday get this, you know, problem under control. I
think it plays into it if we're only thinking about
today. And I don't think that's the best way of
looking at resiliency for these systems. Especially, again, we are
looking at multiple challenges in terms of cybersecurity, physical security
from natural disasters, and people sneaking into these locations and
such. And I think, you know, the natural reality is
that most organizations are just thinking about, you know, the
near term in terms of some of these systems because
the investment is so large. But the investment will be
larger if you have to take your plants or your
grids down for a very long time and you have
constituents who are angry, citizens who are angry, or you
can't produce at all. For some organizations, the cost of
not being able to produce is in the millions per
day. So it's kind of thinking, okay, if we have
to take down our site for 24 hours, that's going
to cost us this much. Plus the cost of changing
everything within there versus being down for potentially two weeks,
three weeks because of some kind of damage, and then
backups or something, or like a natural disaster. So it's
doing some cost benefit analysis. And I don't know if
everybody's necessarily doing it because it's very expensive no matter
what. It's just one much larger price point than the
other. How would everyone feel if they just, for just
the smallest example, lost water pressure for a couple of
hours? People are going to complain about that. Yeah, people
will absolutely. Just water pressure. Forget the lye, forget the
fluoride, forget turning the water off, just lose water pressure.
For a couple of hours and listen to your neighbors
all start coming out into the middle of the street
and talking about the fact that there's no water pressure.
I was going to say, Dave, literally like a year
ago, that's it. That happened in my neighborhood. And you're
right, it's just water pressure. We're not even talking about
something that's like life or death. It's just everyone's suddenly
out in the streets. Is your water pressure working? Is
your water working? Is it just my water? And yeah,
so, you know, that does. You're right. That, that, that
complicates things even more. You know, I guess, I guess
we'll round out the segment by saying that I don't.
I feel bad for the folks who have to make
these decisions because Lord knows I couldn't make them. Let's
move on to the next story. Hackers steal real. Now,
speaking of cyber attacks that have some very material consequences,
proofpoint reported last week on a sophisticated cybercrime ring that's
targeting freight companies to steal their cargo. Now a real
quick summary of how freight companies connect with carriers might
be in order for those who aren't familiar with it,
they tend to use a kind of marketplace system, right?
So a freight company comes in and says, hey, we
have this load. Carriers come and say, hey, we can
carry it for this amount. They make a deal. It
works that way. The way this cyber attack operates is
that first the hackers get in there and they impersonate
a freight company and say they have a fake load,
carrier comes and bids on it. They use that opportunity
to compromise the carrier's account, and then they act as
the carrier and they bid on real loads and say,
hey, we can take this job. When they secure the
job, they send their own trucks over to pick up
the load and take it. And of course, it never
makes it to its final destination. I just, the brazen
ness of this really struck me, which is why I
wanted to kind of bring it up and see how
folks feel about it. Claire, let me start with you.
Any initial reactions to this cyber attack? I kind of
think about it and in terms of like, that's such
a mixed bag where you're just stealing people's Amazon packages.
Like you can either get a computer or you're getting
just soap or something. Where, I mean, I guess with
cargo, it's, I'm, I'm like, initially I was like, oh,
that's interesting. Like, how do you get rid of all
of that cargo that's within? But I thought it was
just an interesting overall concept where it's just like you're
reselling these items and it's. You have to be really
organized to take on that kind of level of crime
and to kind of figure out what you're going to
do with all those shipments as well. When I was
reading kind of the Reddit thread that you sent over,
Matt, about the person whose organization had faced this, it
came down to these people receiving an email saying, you
have to do this for a specific order and then
they input their information. And that's kind of how the
threat actor got in. And it's like, of course, in
my mind, it came down to somebody rushing and clicking
a link they shouldn't have. And I wonder how much
more common this is than reported, because shipping is now
like, everything has to be shipped within two days kind
of thing. So there's probably just a tremendous sense of
pressure on these people that, like, sending them a phishing
email like, that would definitely get them to act. So
I wasn't surprised seeing that. But it's interesting to see
because sometimes I'm guessing, like, people will get packages delayed
and such, and you don't necessarily know about why. And
maybe it's because it was actually just lost by somebody
in a warehouse or it was part of a larger
cargo cyber crime ring. Yeah, it is funny, right? It
is kind of like a scaled up version of the
mail theft ring. I didn't really think about that, but
it's, it's extremely obvious. And also, side note, one of
my proudest moments was foiling some male thieves in my
neighborhood. But we don't have to talk about that right
now. Good for you. You were going to, you were
going to say something. Come on in here. I was.
I feel like I'm watching a true life episode of
Shipping Wars. I don't know if anyone's ever seen the
show. Yes, I have. But that's what this feels like.
It feels like Shipping wars, except it's the old west
version of Shipping wars where they're jumping on trains and
stealing things. But no, like Claire was saying, it's, you
know, when you start talking about whether or not your
package has been lost or whether it's been stolen. I've
actually had packages stolen from the truck, not from my
doorstep. It will say, oh, well, your package has been
delayed, and we'll let you know. If it doesn't arrive
within two days, let us know. Well, so these pirates,
these thieves, these people that do this, I don't understand
what the end goal is. Aside from making money and
making other people miserable. And they do. They make people
miserable. That could be a load of diapers for a
new grandfather, let's say. Or it could be, like Claire
said, a computer. Why are they doing this? Because it's
easy to do. That's why. The attack numbers have increased
year over year. Like 2024, I think it was 25,
27%. And then projected for this year, it's another 22%
increase. I mean, it's. They're going after large transport firms
and it's just a matter of time before they hit
an Amazon truck and then we're all in trouble. Now
I'm sort of wondering too, in light of that last
story that we were talking about, you know, in terms
of the OT systems being attacked, and now we're talking
about these cargo theft getting into the mix. It almost
feels to me like our concept of cyber attack needs
a little bit of updating. And again, maybe this is
just myself and my own projection, but I think cyber
attack and that cyber part comes in pretty big, right?
You still kind of think this stuff is happening on
computers, but it's spilling out into the real world in
these very interesting ways for me. And I just, I'm
wondering if, I don't know, do other people kind of
feel the same thing going on? Or maybe you've been
thinking about this much longer than I have. Austin, let
me, let me throw it to you. Have any thoughts
on this kind of blurring of the lines between real
world security, physical security and cybersecurity? Yeah, that's an interesting
point because, you know, the tides have shifted. You know,
we think of cyber attacks as strictly being digital and
data oriented, whereas these, these attacks are having real world
physical implications. And I don't think people, at least the
general public, fully fully understand that, especially when there's so
many different third parties involved, smaller brokers, multiple carriers within
those supply chains. So that creates much more fragility amongst
those supply chains in this logistics and transportation sector. Absolutely.
Claire, any thoughts on your end? Yeah. In this cyber
range, we've increasingly seen clients go from being more focused
on the loss of data, which is an awful thing
for a lot of US consumers, to more like material
impacts of having either impacting shareholders, employees, and again, customers
in terms of service availability. So I think it's also
a bigger concern for organizations too, where it's like, I
will not be able to produce X products and there
will be an outage for a certain amount of time
and then all of our customers are going to switch
to Y product because it's not available. So I think
as a consumer we also don't care as much about
our data being lost because our data has been lost
how many times. I'm sure everybody on this call has
like, you know, at least 10 years of data credit
monitoring. So you know, like, I think it's something that
organizations are also thinking about because they're scared because that's
more of a, like an impact to their business if
they can't produce or provide services and that's going to
impact them more. And I think at the end of
the day, if you can't kind of going back to
the OT, if you have to pay $2 million to
get your site back up and running and it costs
5 million a day of downtime, you're going to pay
the 2 million in ransom. You might not admit it,
but you're, you're going to pay it because it's, it's
again the cost benefit analysis. So you're either going to
lose X number of customers and it's going to cost
you this much or you know, so I think it's
just kind of a, it's a mental shift a lot
for businesses and then consumers were starting to get annoyed
by it as well as it trickles down. I'm going
to age my 50 year old self here. Back when
I got my first computer back in the early 90s,
we'll go with that. No one ever thought about the
number of IP addresses that we had or had available
or whether or not we were going to run out
as our landscape shifts to having computers control everything. And
let's be honest, it does. Everything is a cyber attack
now. So we can't just think of cyber attack as
one person attacking a computer. You have to put that
the supply chain, that's a cyber attack. You know, the
physical attacks, yeah, those are different but for the most
part we're looking at cyber attacks on everything. So yeah,
our definition definitely needs to be updated for cyber attacks
right now. It's still the old school definition and we
need to get away from that and towards a more
modern definition. Move on to our next story here folks.
This is malicious packages that drop time delayed logic bombs.
Researchers at Socket identified nine malicious nougat packages with hidden
logic bombs designed to detonate years after they've been installed
and working properly the whole time. So these packages, they
do work, they do what they say on the tin,
but hidden inside of them are these logic bombs that
are set to detonate, go off after 2027 or 2028,
depending on what package you're dealing with. And the way
they work, kind of their malicious components are that, you
know, when you first install them, they will start immediately
terminating some processes. But they only do it with a
20% chance because it's designed to kind of make it
seem random, like it's not related to the package, it's
just something crashed. But then the specific kind of code
that detonates after a few years is they start silently
corrupting write functions. Right. And so I just. The thing
for me was trying to figure out what's the point
of the time delay, what does the attacker gain, what's
the play here? And I want to throw to you,
Austin, what do you think about this technique? Well, first
off, I've actually never even heard of this technique being
used. So it's pretty, pretty wild that this is going
on. But as you're asking, what's the point here? Well,
I think it's really about the dependency on time that
because this gives attackers so much dwell time, by the
time say the payload is triggered, that initial dependency has
been forgotten because so much time has passed already. So
attackers really aren't thinking and days, weeks or months, they're
thinking in years. And that's where they're using time to
their advantage and these threats are becoming successful. It's funny,
right? The attackers have the long term thinking, but we
talked quite a bit about how our organizations don't always
have the long term thinking. Right. When it comes to
OT patching. There's a little bit of an asymmetry here
that is, I don't know, something to think about. Dave,
any thoughts on your end about this technique? Yeah, so
what they're looking for is complacency. How many times have
you installed a piece of software and five years later
it's still sitting there? Whether you use it or not
is irrelevant, but it's still sitting there. I personally probably
don't uninstall about 90% of the software that I install.
So if I'd have downloaded one of these NuGet packages
in couple of years, I'm going to have a logic
bomb go off on my machine because I just don't
do it. I don't uninstall things. And I think that's
what these attackers are counting on. They're going to get
these packages installed. They've been downloaded what, 9,500 times. So
9,500 people who haven't read this story are going to
be in for a very big surprise come September of
27 and October of 28th. That's a really good point.
I mean, I'm looking at my own desktop right now.
It is littered with things I've downloaded I haven't touched
in forever. And you're right, I don't bother to get
rid of them. Storage isn't really a problem anymore. So
you download a thing and you just let it sit
there. That's a good point. I hadn't thought about that.
I should delete some apps. Claire, what about you? Any
thoughts on your end? I just think it's smart from
a threat actor perspective. It's just. It's smart. Most people
will forget about things. Most people won't even read that
headline and know that there's something they shouldn't have downloaded
in that. In whatever they downloaded. So I just think
it's really like a innovative way of getting malicious packages
on people's devices without them knowing, even though they, at
this point, I guess, should know since it's out there.
But it's just smart from a threat actor perspective. Yeah.
Austin, I want to come back to you again because
as you had said, right, this is something you hadn't
really seen before. This is kind of a new one
for you. I'm wondering, do we need to be paranoid
about this kind of thing lurking in more packages? Do
we need to be worried about this or do you
think this is one incident? How should we approach this?
How should we think about this is kind of threat
on a scale level? Any thoughts there? Yeah, I mean,
I think we should prioritize this threat because again, the
threat landscape's always evolving and we're seeing this play out.
And so we need to sort of think ahead of
the threat actor because there is some strategic intent here.
This is about long term persistence and also leverage by
using time to their advantage. And again, it's not just
about stealing data. It's about planting those seeds for future
leverage and control as well as just overall sabotage. Gotcha.
Dave, any thoughts on your end about how defenders should
be thinking about this kind of thing? Typically, I would
think that the defenders of databases and industrial controls would
wipe it out completely. I mean, now you've got four
years worth of data that you can play with. Industrial
controls, same thing, you've got four years worth of data
to play with. So you're in for some really big
disruptions if we don't get all of these traced or
as many as we can traced to the downloaders and
get them off of those systems because they're probably going
to extremely important functions. Absolutely. No, that's a very good
point and I neglected to mention, but yeah, these packages
target largely database systems and icss, and so yeah, they
can cause some serious problems there. Let's move on then
to the next topic. And this is honestly maybe the
one I'm most excited about to ask you folks about
reports of AI powered malware that we've seen come out.
And in the last week, two big instances happened that,
that made me want to bring this before you folks
today. The first was that Google started talking about an
experimental malware it found called Prompt Flux, which can interact
with Gemini's API to regularly request rewrites to its code
with the intent of evading detection. But as cybersecurity researcher
Marcus Hutchins points out, this self modification function was commented
out in the code, so it wasn't even actually active.
And even if it were, it probably wouldn't operate all
that seamlessly. Just because it's requesting, you know, code rewrites
doesn't mean those rewrites are actually going to evade any
detection system or they might not even work. The other
thing that happened was that there was a widely cited
MIT Sloan working paper that came out in the spring
and it claimed that 80% of active ransomware attacks were
using AI. It has since been withdrawn after some serious
criticism came out regarding the methodology, with researchers wondering how
the authors were able to definitively say AI was involved
in any of these attacks. So my question is, have
we lost the plot when it comes to these AI
powered threats? You know, are we getting a little ahead
of ourselves with the way we talk about these things?
Dave, what are your thoughts here? So there are built
in protections in, in all the AI models for the
public iterations of these, these AI models that keep you
from putting your code in and getting some more malicious
code out. However, you can take the training wheels off
of these and put your own instance in. And now
you've got your own little personal AI bot who is
willing to do whatever you want. And as long as
we, and we talked about this on the other show,
once you start removing the gates, you start allowing more
things to happen, we're going to see this more and
more, the research paper from MIT notwithstanding. I mean, they
did make some grandiose claims, but you are going to
see AI malware come into existence and it's going to
be dangerous and it's going to be good and it's
going to be a way for adversaries to get into
systems that they wouldn't normally have the skill set to
get into. But, but AI absolutely has the skill set
to do whatever it needs to do, programmed by humans.
It's a lovely partnership. You know Austin, your thoughts on
this kind of subject? Do you think we are maybe
a little sensational about this or do you think this
kind of thing is something to be worried about right
now? I mean, I think AI, the idea of AI
being fully autonomous, self evolving malware is sort of in
the realm of science fiction than reality. Because at the
end of the day, AI is a tool created by
humans and it's really a reflection of us. It's not
independent of, of human nature. Perhaps that day will come.
But I think for now that isn't fully operational. You
know, I don't think AI is going to become conscious
and self aware. So I don't really see a 2001
Space Odyssey scenario playing out. But maybe I'm just being
too optimistic. So. Yeah, yeah. And it's kind of like
Dave said, right? The threat necessarily isn't the sentient malware,
but it's like if a hacker has their own little,
you know, training wheels off model to help them do
things they couldn't do otherwise. Right. It's, it's sort of
less sensational, but just as dangerous, maybe more dangerous. Claire,
your thoughts on this kind of situation? Yeah, I agree
with Austin. I don't think it's something, it is very
sci fi right now. But I do think it would
be a mistake to say that threat actors aren't using
AI in some capacity. I mean, again I mentioned in
like, I think our last episode that I was on,
So they are also trying to optimize their operations, which,
which probably includes bringing in AI to that. So I
mean, it would be kind of silly for a threat
actor to not be using AI to check their code
or to write phishing emails or something. It would just
extremely speed up their operations. So I think it's safe
to assume that most of them are using AI in
some capacity like that. But I don't know if we
fully have gotten to the point of like self evolving
malware that is, is almost like a real like biological
virus of some kind. I. We're just, I don't think
we're there yet. And if we are I don't, I
don't know if we'd pick up on it yet. But
I, I mean, I don't necessarily think it's like something
super true yet, but I think a lot of people
are like, forget that AI can be a little scary
if you are trying to do prompt injections and trying
to evade the safeguards. I agree that I don't think
it's sentient yet. And I don't think that anyone would
be foolish enough to think that AI is writing malware
itself. It's all human based. I think it's more code.
What is the word code embellishment that AI is being
used for looking at the code that, that they're putting
in. Hey, make this better for me. So it's not
the AI writing the malware. It's actually more in the
prompt that the individual is using to get the AI
to make the malware better. So I don't think it's
sentient either. I don't think we're going to have a
HAL 9000 situation or a Skynet situation come up anytime
and probably my lifetime anyway. But they are using it
to make malware better because they do have these sandbox
AI instances that allow them to do things that they
wouldn't normally be able to do on say, Gemini or
Copilot or chatgpt or anything else like that. Yeah. And
I want to ask about something else that I've seen
cybersecurity researchers increasingly talk about, is that some of this
might be a kind of communication issue. And by that
they mean some of the organizations reporting on this malware.
They might feel a little, I don't know, motivated to
make it seem more impressive than it actually is for,
you know, first of all to get some headlines, to
get some clicks. But also because if you're helping to
kind of secure AI systems, you might have an investment
in making the AI systems look super, super spooky. So
I'm kind of wondering if you folks feel like we
might be dealing with partially a communication issue here. Are
we just like sensationalizing things too much? Is that. What
do you think is going on? And I'll start with
you, Austin. What do you think about that? This recent
report that was what redacted it has definitely been sensationalized.
Because if a board believes, you know, AI is driving
80% of, of ransomware, which the stat they put on
the headline, you know, they may start overspending and misallocating
resources on AI detection tools, which will also underspend on
basic security hygiene like Patching, segmentation and identity and access
management. So I think that misallocation of resources can become
a massive strategic risk for organizations. Cool. All right, well,
then we will wrap that section up and move on
to our final story of the week. Oh, my favorite.
The password for the Louvre's video surveillance system was Louvre.
By now, we've all heard about the theft of some
$100 million worth of jewels from the Louvre. And in
the wake of this incident, some news outlets have resurfaced
some old reports on the Louvre's security measures, including warnings
from the National Cybersecurity Agency about its kind of weak
systems and the fact that at one time, at least,
the museum's surveillance system password was Louvre. Now, Dave, given
that you were so eager to dive into this one,
I feel like I should just throw it to you
first. Give me your thoughts here. What do you think?
What do you think were they thinking? Honestly, password 1,
2, 3 would have been better, you know, and I'm
sure they probably tried that and said, wait a minute,
let's just go with the most simple thing and let's
just see if it's luv. And they got in with
that. And this was the camera system, right? This wasn't
like the password to the Louvre cyber security system. It
wasn't. I'm sorry, their security system, it wasn't keeping the
jewels safe. It's just the camera system. Just the camera
system? Really? I mean, that's a pretty big system there.
I mean, this is a building that houses hundreds of
billions of dollars worth of priceless. I mean, you can't
put a price on it of priceless items. And they
couldn't be bothered to put multi factor authentication in there.
They couldn't be bothered to come up with something was
more than six characters long. They couldn't be bothered to
put even an exclamation mark at the end of it.
Really? Or. You say for the E speak in there?
Yeah, exactly. We need leet speak in there so that
at least someone with half a brain would have to
use it. Nope. Nope. My son could have probably done
this and he doesn't know anything about computers. Claire, I
wanted to ask you as somebody who, you know, I
know you spend a lot of time working with organizations,
training. I'm sure the Louvre is not the only institution
that's been guilty of these kinds of lax sort of
security processes. And I'm not asking you to tell us
about other ones you've seen, but I'm wondering in general,
have you noticed these kinds of lax Measures in places
before. Is this a trend? I think it's pretty common,
but a lot of people will not admit to it.
They won't admit that their password is admin or password
or whatever the username is, but I think it's very
common. But a lot of the people that we have
that come through the range probably wouldn't admit that. And
these are, you know, executives, so they're not always like
the most tech savvy folks either. They pro. They probably
will not admit that that's their password as well. This
is just, I feel like the password and the surveillance
camera thing, that's like one of the first things you,
you do when in a heist movie is you disable
the camera. So it's like, even if the security association
or the national whatever coalition it was didn't say you
should do something about this, if you watch any of
those kind of movies, wouldn't, you know. But it also,
to me, this whole thing makes the heist so much
less impressive because, I mean, just seeing all of the,
like, lax security measures they had, it's like I. Someone
could have a lot. Anyone could have done it. And
I don't want to say anyone could have done it
because it's definitely something that's very difficult. But if you
look at like the Isabella Stewart, Stewart Garden, like heist
in the 90s in, in Boston, that was just like,
so well thought out. They still don't know who did
it. But it's like with this one, they just kind
of went in and they like typed in Louvre and
called it a day. And they like, they used their
drill and they took the jewels and walked, walked away.
So. And we didn't find them yet. So it's as
far as I know, I think there's suspects in. But
still, I believe they're still at large. But yeah, they
just kind of wore construction uniforms and walked in and
nobody really stopped them until they were taking jewels and
at which point it was over, you know. Austin, any
thoughts on your end? This was a matter of not
being a technical flaw, but it was a flaw in
human behavior and governance. I mean, that's really the big
failure there. Sadly, convenience typically overrides policy when it comes
to cybersecurity. And without any accountability, those critical systems remained
pretty defenseless, especially with CCTV cameras. So kind of crazy
that, you know, an institution as big as the Louvre,
world renowned, thought maybe they were too good for any
password hygiene or password rotation. Also, like, how many people
pass through the Louvre a day? I'm sure. There are
other devices that are just available that you could type
that password into and get in. Like, how many people
are in there that. That, you know, could. Something's not
watched and you could just type it in. It's scary.
You know, I want to end on a constructive note
here. So I'm wondering, you look at a situation like
this, what kind of lessons you think we can learn
from. From the lose, you know, maybe lax approach to
some of their cyber security measures. Let's start with you,
Claire. Any thoughts there? There's just so many things that
they could have. Could have done. I feel like training
of their staff as well would have helped too. But
I. I think this whole thing goes back to when
we earlier said, or Dave earlier said, like, everything is
cyber now. So, like, this wasn't a cyber attack. It
was a physical attack, but there is cyber involved. So
it's just like, I don't know, have your folks trained,
have better passwords. I, you know, don't let people just
drive up to your institution with, like, a van with
a ladder on it and go through the window. I
mean, it's just. Claire has ruined every heist movie I've
ever seen now. Ocean's Eleven will never be the same.
I'm sorry. True. I was just gonna say, you wear
a shiny vest and carry a clipboard into anywhere and
no one bats an eye. Right. So it really comes
down to the human layer as. As the first line
of defense there. I like that. I like that it
comes out of the human layer. I also just want
to say real quick, I do think not letting people
drive up to your. Anything with a ladder is a
good, you know, thing to do in general. Like you
said, Claire, really great advice to Dave. Any thoughts on
your end in terms of lessons we can learn here?
The typical answer is practice good password hygiene. I can
almost assure you that whomever installed the cameras at the
Louvre put the Louvre in as password, and no one
ever thought to go back and change it, least of
all the people who were in charge of it. They
just thought, we'll get to that. And getting to that
was the next day, and then it was the day
after, and then it was the day after, and I
think it just became forgotten. I don't think that the
Louvre purposely used the easiest password that they could find.
I think it was just forgotten. And I think they
probably learned their lesson and it cost them a few
bucks. I think that's a really good place to end
it on. And it is all the time we have
for today, folks. So I want to thank you, Dave,
Austin, Claire, for sitting here and having this conversation with
me. Thank you to the viewers and the listeners. As
always, folks, subscribe to security intelligence wherever podcasts are found.
Stay safe out there. And I don't know, just change
your passwords up, man.