KOPYA NG SALIN 🔗
WEBVTT
1
00:00:00.000 --> 00:00:25.000
How are you going. When we get into regulated environments with agents, how are we going to attribute back if it does something that breaks or violates a compliance or a regulatory issue or a privacy issue, how do we have attribution back to that AI agent? Which AI agent did it? How many AI agents do you have? You know, how are you controlling this stuff?
2
00:00:25.000 --> 00:01:26.000
You're listening to PrOTect IT All, where Aaron Crow expands the conversation beyond just OT delving into the interconnected worlds of IT and OT cybersecurity. Get ready for essential strategies and insights. Here's your host, Aaron Crow All right, everybody, thank you for joining me for another episode of the PrOTect IT All podcast. One of the, one of the joys of this thing is the people that I get to meet. I know I talk about this all the time, but having people like Ken on the call, you know, Ken and I know each other. We've shot guns and drank bourbon and, and been at conferences and all the things. So excited about these conversations because at end of the day, the whole point of this is to talk about the things that we're all hearing and experiencing, but sometimes we're not in the same room as, as one another, but we all value from the, the experiences and the knowledge that, that we have. So, Ken, thank you for joining me again today. This is not your first time on the podcast, so for those of you that don't know, why don't you introduce who you are a little bit about your background, et cetera, and then we'll just dive in, man.
3
00:01:26.000 --> 00:02:42.000
Yeah. Appreciate you having me back on again, Aaron. It's been, been too long. You know, I've been, I've been doing, you know, Ken Foster been doing cybersecurity and IT for a little over 30 years now. So I kind of started out in the, started out doing this stuff back in when I was in the Navy before they even had, they even had a rate to do this or was kind of like, hey, you work on these computers that run missile radar systems. Why don't you help us figure out the information security back or information assurance as we called it back in those days and build the networks and do things like that, you know, and it kind of realized that I had a passion for it and moved, moved into the corporate world doing this over the years and been through, been through most industries out there. You know, I've worked in global manufacturing, global fintech. I've worked in the technology startup space. I've done some field CISO consulting for other technology companies. You know, ran grc have kind of been, you know, run government programs as well in and out of the government service. Worked for the intel community as a contractor for a little while doing some things overseas. So it's, I, I think I'm starting to sound like the state farm commercial. I know some things because I've seen some things.
4
00:02:42.000 --> 00:02:44.000
Yep, yep.
5
00:02:44.000 --> 00:03:06.000
Yeah. So. But now it's. And, and I tell you right now we're in such an exciting time right now. I think I'm, I am reading more right now, doing more research, doing more hands on stuff than I've done in a long time. Just because it's such an exciting, I will say hockey stick turn and the way that we're going to be approaching cyber security going forward. So.
6
00:03:06.000 --> 00:04:17.000
Yeah, yeah man, it's, it's, it's, it's definitely a different place than, and it reminds me to your point, right. It reminds me of my early career when I was, you know, reading technical books and building in my lab and, and doing all the hands on stuff and you know, a lot of my career towards the end here. And I don't mean towards the end of my career, I just mean the last few years of my career has been spent more, you know, policy and conversations and architecture and high level thought which is all value add. But to your point, like in the last year I've been doing more sandbox architecture actually hands on building building applications. I've never been a developer, never claimed to be a developer, but you don't have to be a developer now to build things like that. That. Now am I going to build a product like CrowdStrike? No, I'm not pretending I can. But I can build things that can be tools in my toolbox to help me in my job. And sometimes that's you know, just you know, going through a PowerPoint deck and finding all the fonts or hidden characters to building a dashboard. Like I really see AI and some of the things that we're doing now as, as the spreadsheet of the 2000s,
7
00:04:17.000 --> 00:11:29.000
it is and I mean it's, you know I, I've built multiple agents. Some of the consulting work I've done where it's, it's, you know, again outside of the policy stuff, this is actual like work work, hands on kind of architecture work, kind of revamping network and security stuff. And you're like, you know what, I don't need to sit down and enter and this is the where AI I think is really changing what we do. I don't have to sit down. I'm with the, the guy who's been, or the guy that left the company in the case of the job I was doing, trying to figure out what he did, why he did, and what do I need to understand about the environment. I can sit down now and write a purpose built agent from an AI to go in and parse all this data, parse all of the logs, look at the configurations. I can say, okay, now let's take that and now let's measure it against their policy and their standards, or let's measure it against the known. You know, let's measure it against nist, let's measure it against cis, let's see how all this is configured. Because I can now accelerate the time, time that it takes me to do something like that. Whereas before, if I had to manually sat down and plow through a year's worth of logs to figure out what's going on or configuration files, it'd take me months to do that. I think. It took me once I wrote the agent, which took me three or four days
Load full transcript
to get it tuned and get it right the way I wanted and get the outcome to come out the way I wanted. I, it did the work in three and a half hours, you know, so now you've, I think that speaking, you know, the AI thing is, is that you still have to make sure you understand what that outcome and output is. And you still got to check your work because. Yep, look at Claude. What was 4, 4.6 opus was the last one to come out. When it first came out with cowork and all that, it was phenomenal. Everything was great. But you, you see degradation over time, right. All these models, because they get access to bad data because of the way they're designed, there's going to be a market degradation after they've been out for a while. And then, you know, you gotta pay attention to what you know. I think every few days you get a pop up that says, hey, you need to relaunch and update your agent or update the tool or update the model. And Claude, this week, one of their updates gives it more Internet access. So if you just clicked update and didn't pay attention to it, there's some things you need to configure in that because you could be inadvertently sharing your data. Yep. You know, and you know, we're seeing these AI attacks. I mean, I know there was, there was a big crypto attack this year or this week. There was a, there's a router attack that come from Russia that they didn't, it bypassed mfa it used tokens and it targeted at home routers. So if you've got, and we still have a lot of people working remote or at least doing a lot of critical work off of their home networks and these AI tools are being, are able to find this stuff so fast, exfiltrate data and do this at a pace that we're just as humans are not able to keep up with. And I mean I, I know Kevin Mandy talked about it at RSA this year. He's like, you know, it's a, AI arms, AI to AI battle. And it's true. I still think we've got some things we've got to learn about it and figure out. So there's still caution that needs to happen as we look and move forward. But you cannot afford not to be looking forward and moving out, moving out of how you're going to, how are you going to implement this in your environment? How are you going to put the right guardrails around, what's safe for you, what's not? And that's another interesting thing. I think, I think the AI boom that we're seeing, which is phenomenal, it's going to change the way we do things, but it's also highlighted. And I had a breakfast conversation with a bunch of CISOs and we were sitting around and we were talking about AI and a couple of things came out very quickly. A, we still have problems with that. We still have an education, understanding problem of what AI agents are and what is process automation. Right. Because an agent's not process automation. If you're asking, if you're building automation to take data from step A to B to C and not really do anything else other than just move or do data entry, that's not an agent right now. An agent, an agent's actually, you're designing it to go out and make decisions and make change, you know, do things like that. So there's, there's an education problem we're still seeing across even the cybersecurity leaders or cybersecurity industry. And then it's also exposed how immature a lot of companies are. And this is, this was across the gamut of large enterprises, large global companies, mid market companies, technology companies. It wasn't unique to anybody, but it was, you know, we still, after 30 plus years of doing this, we still suck at identity and access control, we still suck at data segmentation, we still have a hard time keeping data where it belongs. So there's a data quality issue as well. So what is exposed is how many gaps still exist in very mature programs from the outside looking in on data governance and identity and access management, which is identity, just call it identity governance. Right. Those two things are so critical to a successful and lower risk AI program. So when you're thinking about this, you'd better have those things online. And it's also. So I think Zero Trust is getting ready to become a big hot topic again. We're going to start hearing about it again. I think anything with identity is going to be hot. We're going to see data governance programs picking up. I think the DSPM market's going to really take a boom even more. So we're going to start seeing those, those areas really pick back up as topics of conversation because you need to have very good programs around that to protect yourself. And then I also think it's going to bleed a little bit into the API non human identity. Well, definitely the non human identity, but those two are going to bleed in together as well. Because once you got MCP server set up and it's connecting to APIs and you don't know where it's connecting to you, you're starting to get data moving very quickly to a lot of different places. So it's something that you've got to be putting, you've got to be putting a concerted effort into learning all that you've got. Going back to the API thing, you know, a few years ago when API security was the big hot topic of the, of the, you know, it was average. Everybody goes, oh, I've got a couple hundred APIs. And then you do a discovery and go, oh well, yeah, you've got 3,000 APIs. Right. And that's the same with AI agents. Right. She's like, oh, well, we maybe have a, we maybe have a hundred or thirty agents. Yeah, probably, realistically, probably got 300 to 400 agents. Because what's happening is, is the Microsofts of the world, The Googles, the AWS's, the Palo, CrowdStrike, everybody, they're embedding AI agents into their tools, they're turning them on. And if you don't have a program and understand that stuff's getting added all the time, you now have this unmitigated army of agents running around inside your environment.
8
00:11:29.000 --> 00:11:30.000
Correct?
9
00:11:30.000 --> 00:11:31.000
Yeah.
10
00:11:30.000 --> 00:12:02.000
And it's going to, you know, if you really look at. I just read an article yesterday, I saw your post on the Mythos thing we talked about before recorded that OpenAI has released their version of the same thing, you know, for years. Obviously on the OT side, you've always heard this thing of, oh, you guys have really old technology. You're not patching as often. Like you guys are less secure than the IT side because we're patching, we're on the latest revision of the firmware, et cetera, et cetera, et cetera. We just normalize that. Like now with Mythos, your brand new system is no different than Windows xp.
11
00:12:02.000 --> 00:16:10.000
Oh, it's not. And, and to be honest with you, to be honest with you, there's probably a little bit of, and I don't want to be flipping about it because it actually has been true is sometimes if you're not patched up to the latest patch, you've actually protected yourself against some of the zero days that are coming out, but that AI is not able to find. But again, it comes into segmentation. Right. If you can't patch and you can't keep it up to date, then your, your biggest defense is going to become network segmentation, good identity and access control because you're going to limit who has access to it and who can get access to it and those OT environments. And you know, I've ran been in charge of the infrastructure for some pretty large ones as well in, in the manufacturing space and you know, making sure that the plants are isolated, that they only have access to what they need and only who needs to have access to it is going to become critical. Because you're right, I mean, Patch Tuesday was this week. Microsoft, I posted on that too, 168 vulnerabilities in patch Tuesday. And one of them was a major sharepoint flaw that's been around for a while. Yeah. And it was pre Mythos. Right. One Tuesday. Right. And if you're not, if you're not, and that's very difficult in a large enterprise and when you're dealing with thousands or and maybe even hundreds of thousands of endpoints that need to be patched and updated and then you've still got all this other work you've got to do. If you don't have a mature vulnerability management program that's risk based and you have the ability to prioritize and you have segmentation and everything else in place that allows you to maybe isolate things that you can't patch on a regular basis. That's you've really got to make sure your programs are very specific, very mature about that. The good news is there's a lot of stuff out there nowadays that can help you do that and AI can help you build those programs as well because you can, you can use it As a research tool, just validate your work, right? That's the, that's the one thing I always caution everybody against, is, yeah, AI is a great tool. But it's. I've said this before, right? The problem with automation is, is automation is great, but automating a bad process makes the bad process worse fast. Same thing with an AI agent, right? If the process was never good, designed, well, to be automated, it's probably not going to be designed, it's probably not going to run well in AI and it's probably going to come with errors in the outcome. And I think this gets us into the. Where everybody's like, oh, we're going to lose all our jobs because of AI, right? We're changing our jobs, we're changing our roles, we're changing. Now, are there types of work, and not only in the cybersecurity industry across the globe, are there things that AI agents are probably going to replace from somebody's everyday job? Yes. Is it going to be instantaneous? No. Because I had this conversation with some sales folks and they were like, oh, we can, we can replace our entire sales operation team with AI agents. 90% of our people can go away. I was like, okay, let's look at your process. Because your process is so convoluted. Because, you know, and it's typical with any sales team I've ever worked with is somebody comes up with all these little different ways of compensating and how they approach sales, how they measure metrics. And it's unique to that. That company, right? It's always unique to that company. And then you go, well, the people who've been doing this and helped build it barely understands how all this works. And you're constantly having issues with it. How are you going to write that into an algorithm, into a program and help it figure it out? Because it's going to decide this is inefficient, we're going to do it a different way, right? It's not going to listen to you and it's going to change your process. So if you're not prepared for that and understand that it's going to take 18 to 24 months probably to figure out the right process for an AI. Now you have an opportunity to retrain, relearn, upskill, to say, okay, your job now is dealing with this stuff that we can't write an AI agent for or that we would prefer it be human touch at that point.
12
00:16:10.000 --> 00:17:13.000
So I'm a human in a loop too, right? And it reminds me of the Old was it I Love Lucy where she's on the manufacturing and it's going too fast and she just starts eating the chocolates or whatever because the process is broken. It gets out of control. And what do you do? So I can. That's what I thought about as you were talking about that. Right. Is. And it's, it's, it's true. Because, you know, it's very easy to use and it's why, like, so a lot of the times when I'm using AI and the local agents and the chat bots that I have and things like that, when I'm giving it tasks, I'm. I'm making it ask me almost every step, especially in the beginning, I'm going to do this. Are you okay with that? Yes. Like, it doesn't exactly save me much time, but I'm fixing my process. Like, no, that didn't work out. Let's not do that. Let's skip that. But I don't want to just say go and let it go run off because it can make mistakes. And then you don't even know what it did. Like all of a sudden it's changed green to blue
and up is down and you don't even know why or how or where it did it. And now you're backtracking from all those things. Right. And I'm doing it in a safe world, in a, in a sandbox, accounting. Just unleashing it to production in your enterprise environment.
13
00:17:13.000 --> 00:17:40.000
This reminds me, I was using it the other day to update a presentation and I went through and made a bunch of updates and I looked at it on the screen. It had mirrored everything. So everything was mirrored. I was like, what happened? I was like, oh, what happened there? And it, and even it didn't know what happened. It just said, let me reprocess it. Because I was like, hey, this is mirrored. What happened? And they're like, uh, don't know. But here, let me rerun it. And then it came out correctly.
14
00:17:40.000 --> 00:17:45.000
And sometimes I think you have to talk to it like your kid. Like, hey, stop being an idiot. Quit doing that.
15
00:17:45.000 --> 00:18:39.000
I read, I read an interesting article earlier this year on should you be polite to AI? And it was a politeness and compared to the efficacy of the output, what its accuracy was, it actually was more accurate now, not just greatly, but, you know, two or three points. Like it might have been 85% accurate being super polite to it. And it may have been 85.6% more accurate by being mean to it. Yeah, like berating it Telling it it's an idiot, and if it could figure out how to do its job, you wouldn't have to yell at it or whatever. And then it. But so it actually was a little more accurate if you were mean to it. That's funny. Yeah, it's. I was like, that's interesting. And I was like, who thought of that too? Who sat down? What if I get mean with this thing?
16
00:18:39.000 --> 00:19:46.000
Yeah, they were in a normal prompt. And I've done it before, where you're talking like you're just responding, don't do this, don't do that. And then finally you get to a place where you've especially early on, where you've told it to do something five times, and it just keeps trying the same thing and beats his head against the wall. Like, a lot of times I'll be building in the lab and I'll be using in a disconnected way. I'll be using ChatGPT or Claude or whatever it is to like, hey, I want to build this, this Ubuntu server, and I want to have this type of capability. So it's giving me commands. I'm then pasting those commands in the command line. I'm not giving it direct access, so I'm, I'm actually pasting it. And then, you know, we're troubleshooting something, like networking's not working or whatever the thing is. And I'm, I'm trying to get it to see how good it is on how it can troubleshoot, and it just keeps trying the same thing over and over again. And I'm like, no, that doesn't work. No, that doesn't work. We already tried that. We already tried, remember? We already tried that. Finally. Sometimes I'm just like, quit effing trying that, you freaking idiot. We've already done that. And then it's like,
okay. And it doesn't try it again. It's like, why did that work? Like I told you five times, you've already done this. When I called you an idiot and started cursing at you, then you stopped using it and you tried something else.
17
00:19:46.000 --> 00:20:21.000
Yeah, and that's the, that's the interesting thing too, right, is there's a big difference between running it. Like, you and I do the same thing, right? I'm like, here's some code. Let's. Let's build a Python script to do X. And I have it build me the script, I copy and paste it over to my sandbox that I've got. I do some, see if it works, and inevitably you're Going to have some little tweaks to it because it's doing it based off of generic data for the most part. Right. So now you may have to change some specific paths or stuff like that. And that's fine because I'd rather again not give it direct access to that data.
18
00:20:21.000 --> 00:20:22.000
Correct.
19
00:20:22.000 --> 00:21:43.000
And doing that, I just want it to help me write a, a more efficient code. Right. Or a more efficient PowerShell. I know I can do this in PowerShell, but it's going to take me 20 minutes to sit down and get it all typed out because I got to research it a little bit where I can tell it, hey, I want a PowerShell script that does this. Give me the framework for it and then I can dial it in for my thing. And I think that's the part of the, where that shifts. And what people are going to be learning about AI and how their roles are going to be, you're going to know based on your experience, based on your history in the job that you do, you're going to know how to ask it and how to validate the data. So that's where it's going to make the change in what your role is. Your role is going to be more of a teacher. I mean, the best way to put it, you're going to be a teacher for your AI coworker. Right? Because it is going to be your co worker. You know, I mean, Claude called their stuff cowork for a purpose, but I mean, it really is a AI coworker to an AI entity, just like a worker in your environment. Right. So in a lot of ways you've got to treat. And here, here's something that I have heard very little talked about and I've talked about it a little bit, but how are you going. When we get into regulated environments with agents, how are we going to attribute back if it does something that breaks or violates a compliance or regulatory issue.
20
00:21:43.000 --> 00:21:44.000
Right.
21
00:21:43.000 --> 00:22:38.000
Or a privacy issue? How do we. Do we have attribution back to that AI agent, which AI agent did it? How many AI agents do you have? You know, how are you controlling this stuff? So you're going to have to have. Every one of them are going to have to have an unique individual identity with access control, because in a, especially in a regulated environment, because you're going to have to be able to attribute back. Now the question is, is it the agent's fault or is it the person who wrote the prompt?
Right. Is it the person who wrote the. Wrote the agent? So who, who ultimately is responsible for if it does something malicious or does something violates regulatory or a law. Right, because that's, that's a question that you're going to have to ask because the auditors are going to figure it out before too long. They're going to want to know that answer and they're going to want to understand, hey, how do you track these things? How do you keep up with them? And then you got.
22
00:22:38.000 --> 00:23:15.000
Ultimately, ultimately I think that's where we're going to want and I can see it in some of the regulated environments. Why we're not going to use those AI agents because of that. Right. Until I can know that attribution, that documentation, what is an auditor going to look at? How do I prove that? How do I, how do I show that? How do I keep it in its, in its. Just like, you know, in nercsep, we don't use, Nothing's in the cloud, everything's on premises. Right. Same thing. Like I, I don't see us using AI in a regulated environment, especially in this type, in that kind of regulated environment because they're not, it's too much of a risk because it can do things that. It broke my compliance. Now I have a fine because AI did something stupid.
23
00:23:15.000 --> 00:24:19.000
Yeah. And, and that is, that is a real, real threat that's coming out there. And it's, it's interesting too, just speaking like NERP sips, there's a report that just come out on CMMC readiness for the government. Right. And they, they looked at, and right now CMMC is still focused. You have to be a defense contractor. Right. It's still, they're still focusing it on defense, but from the report, only 18% of the people that they reviewed of these, these 100 plus defense contractors, we're able to demonstrate readiness across the foundational cmc, CMMC areas. Right, right. It's a pretty low number that are ready for CMMC, you know, and only. And it also had only 28% completed and documented their internal SPRs assessment. Right. So that's basically your risk assessment. And 67% haven't confirmed that their cloud vendors are FedRamp.
24
00:24:19.000 --> 00:24:20.000
Right, right.
25
00:24:20.000 --> 00:26:08.000
So that, and these are, you know, highly, highly regulated. I've ran These FISMA and FedRAMP environments for companies and you know, that's a, that's a huge concern because, and again it, I do know it all depends on the agency because some agencies I was working with were way more mature and way more concerned about the actual vulnerability management and just how the program ran. You had other agencies that all they cared about was uptime. They could care less about keeping it patched and up to date all they wanted, because they wanted it serving its purpose and serving it to the customers. So all they really cared about was can we keep it up and can we fail it over and it come back up instantly. So. And because they were allowed to operate that way, that's where
CMMC is going to change a lot of this with them being forced to do that because now they have to focus on actual vulnerability management, actual risk controls and locking stuff down. I think, again, this goes back to our conversation about OTI and OT networks as well. You're going to have to isolate stuff because you're not going to be able to patch some of this stuff or give a new version of it. The medical industry is the same way. Right. If they have to have a certified health device from the government for their, that process takes forever to get it actually certified. So if you have to change a piece of hardware or do something that requires that certification process, you're kind of stuck at the speed of the government. And we all know how that typically goes. It's not the fastest thing on the planet. Right. So. So then you got to put in a lot of compensating and a lot of mitigating controls out there to figure out how to isolate and protect this stuff.
26
00:26:08.000 --> 00:27:15.000
Yeah, I mean, it comes down to, you know, really looking at, it's going to change the way that we really focus. Right. So we've got a. There's a lot of great tools out there. And I don't want, you know, I was a vendor, I was a CTO of a software company. Like, I get it, there's great, there's great capabilities and products that are out there. But this is going to change the game. It doesn't mean that those things are invaluable, but it does mean that, that we, we have to change the way we protect our environments. Right. So we have to do more defense in depth. We have to be able to, you know, we're. We talk about identity of a person, but we're going to have to more a process, an endpoint. Like all of the layers of osi, we have to have identity on all of those things, not just one of them. Because as these AI agents grow, we have more and more of them. We don't know who's doing what. So we need to be able to understand and identify does this process, does it need to be able to talk to that thing and who is the authority to give it granted those rights because Aaron thinks it should have rights. And Aaron just happens to be in the right domain. Admin group or whatever group gave me access to it. Now I'm doing something that I probably shouldn't be doing, but I just happen to have it. And organizations are going to have a long time to figure that out.
27
00:27:15.000 --> 00:30:49.000
Yeah. Also what you're looking at, right, is the fact that this is a real world story. I won't name the company, but it was talking to a plant had back off office it, plant it. Right. And we were talking about network segmentation and how to lay that out. And the plant it, the head of plant, it said, well, the plants don't ever need to talk to each other so we can just wall them off from a network segmentation. Plants are self contained units. They should never be talk. Yeah, they're air gap Oak Bay. So made some network changes to start air gapping the plants and immediately we got tickets put in by the plants saying they couldn't talk to the servers in the other plant. And it's like, okay, so you guys don't understand your environment well enough to understand the plants must be able to talk to each other. Same conversation about Internet access. Like, oh, we don't need Internet. We really don't need Internet access to keep a plan up. And then we find out their material handling system is a SAS product that they bought that they can't unload a truck and put the rolls of metal in the right place in the plant without a barcode scan that has to go to the cloud to do that because it's not on prem. And that is. That's a great analogy, I think of what you deal with not only in that environment, but you deal with it with software development internally too. Rarely do you have somebody at that does this piece of the software stack that understands how all of it works together. So I think we're going. It's also this whole. And this is the way I like to frame it into cyber resiliency. Right. Because cyber resiliency is not about stopping the attack. Resiliency. And I don't even like calling it cyber resiliency. It is business resiliency. Right. And it's just cyber. Cyber's part of that, just like it is part of it. Just like every payroll AP is all part of the resiliency of the business. Right. What you need to understand is how all that works together and how fast can I get it back up? That's resiliency. When we get hit, how fast can we get it back up. Does that mean we fell over to another product? Does that mean we have secondary and tertiary, depending on how critical it is to the business? And then what is the budget and the cost that goes along with all that? And those are discussions that are rarely had with the right people in the room. It's not that they're not happening. They're not happening with the right people in the room. Because you're going to have to find that person who understands how that pro. Just like us talk about with the plant, finding out that they have a material handling that they have to have up and running to be able to actually run the presses in the plant. And nobody else was aware of that, right? Except for a few people. And it's like, okay, guys, this stuff has got to be communicated broadly. Same with AI, right? As we're writing agents, they have to be documented. We have to understand what the intention is, what the outcome of that is going to be. The answer cannot be. And I had a CTO tell me this not long ago, we want to give AI to everybody and just see what happens. And I'm like, you've got to be shitting me. I was like, you don't. That is the. You know, I didn't say it out loud, but in my mind I was going, that is the dumbest thing I have ever heard. Because what's going to happen? I'm okay, if we build a sandbox isolated, give them some dummy data and go, hey, you guys spend X amount of hours a week in this sandbox environment testing out your theory?
28
00:30:49.000 --> 00:30:50.000
Sure.
29
00:30:49.000 --> 00:33:26.000
Not in production, not in. Not in the primary network. Because it goes back to the. And I've quoted this before, but you remember the movie Sahara with. And there's that one story in there where he's talking about the coin that he finds in his grand and his dad's closet in a shoebox. Coins from somewhere overseas. And it's from, you know, 1400s or something like that. And they go, what's the point of the story? He's like, well, coins travel. It's the same with data, right? Money travels. Data does, too, because data is money, right? So data travels because your HR department, your AP department, your salespeople, every one of them have downloaded data and put it into the most comprehensive business tool out there, Excel. And it lives in an Excel spreadsheet somewhere on their environment. And they. And it's probably become so critical to their process, they have 14 copies of it spread across your network, across their desktops, across other desktops. Across possibly their home networks. Right. So and now you can embed AI into Outlook or into Excel. Right. So now Excel can have a CLAUDE agent or a copilot agent built into it. Now you can tell it to go out and do all this data manipulation and go find all these other Excel spreadsheets. So the next thing you know, it's connecting to all these and it's data that you have almost zero control over from a security standpoint because you've allowed it to get because of over the years and before things were segmented off, hopefully, or you've decided to lock down the way people do their job, this data had migrated out. Now here you. And that creates another problem. Not only is it risk from a data exposure, they're probably using stale data, very likely that old out of date data, which is a data quality problem. And I think you really look at AI when we talk AI and two governance buckets. There's a security and risk bucket, but then there's a data quality business outcome bucket that is very different from a security and risk. Although there's some access controls and stuff like that do interplay together from that. But in reality is how stale is this data? Am I, am I so married to this agent that I spent all this time writing on that I haven't updated it or looked at the new things are coming on? And have I made the decision to keep running on this old version instead of having it rebuilt or redo it? Yeah, and that's, that's stuff that people have got to be thinking about. And then the other thing is budget.
30
00:33:26.000 --> 00:33:27.000
Yeah.
31
00:33:27.000 --> 00:33:57.000
If I've got a bunch of stale agents that are just sitting out there and still running and they're eating up tokens, they're eating up compute, they're eating up storage. If I'm running them in my cloud environments, a big enterprise that can become a very significant amount of money very quickly, that these agents are just out there churning away, burning through tokens and next thing you know you fire. You've had let your sales department build a bunch of AI agents and then you get your, then you get your AWS bill and you're like, what the hell?
32
00:33:57.000 --> 00:34:07.000
It's like, it's like that back in the day when the kids would, you know, were on the phone and they were playing the apps and they were buying tokens or whatever and then all of a sudden you get a 500 cell phone bill like what the hell happened?
33
00:34:07.000 --> 00:34:08.000
Right?
34
00:34:07.000 --> 00:34:08.000
Yeah.
35
00:34:08.000 --> 00:34:14.000
In this case you may be getting a five million dollar bill. Right. Or more Right, Yeah, yeah.
36
00:34:14.000 --> 00:34:57.000
And they'll let that slide once, maybe, but not, not again. But yeah. I mean, ultimately, this is where, and I talk about this all the time, like it really gets back to all of these things we're going to. I don't see us going more advanced in the beginning. I see us really going back to the foundations, documenting your processes really well, understanding the things that you need to do, understanding where my risks are, how do I recover. Doing tabletop exercises, like real tabletop exercises, not the ones that a lot of organizations do to check the box with 12 people so they can get lunch. Like real things and do them more, more frequently. Going through more scenarios. Because one tabletop, you're talking about one scenario that's enough to do once a year.
37
00:34:57.000 --> 00:34:58.000
Yeah.
38
00:34:58.000 --> 00:35:02.000
It's something you should be doing consistently with your teams and multiple scenarios.
39
00:35:02.000 --> 00:35:09.000
Tabletops, once a year. Tabletops are a compliance checkbox. I mean, I, I'm, I'm, I think they're a waste of time if that's the way you're approaching.
40
00:35:09.000 --> 00:35:10.000
They are.
41
00:35:10.000 --> 00:37:51.000
Yeah, they're, I think, complete waste of time. Gets the, gets the auditor off your back. Right. But if you really want to use them to understand the gaps, understand the, the problems that you have with your strategy, your program, understand and help the business leaders understand why what you do is so important to what they do and help them understand it, they've got to be involved. You've got to get the right people in the room. And I say this the same thing. If you're looking at a new technology, let's say we're deciding between Claude, Gemini, Microsoft, bringing those in to look at them for enterprise. And I've just got the cybersecurity team, or I just got the IT team, or I've just got the development team in the room looking at this, you're now putting a much longer Runway on, because now you're going to have to go back and bring that team in and have a conversation. You're going to have to bring the other team. You got to get the right people in the room and have that discussion. Before you ever invite the company in to have that discussion, you need to make sure you guys are all on the same page. You
all understand what that process is, how we're going to roll this out, what are our, what are we expecting it to do for us? What are we, how are we going to test our outcome? What does good look like when this does what it's supposed to do? And then you bring them in and you have them Explain their products to you and then you go down that path and do that. It's see that so many times where you and security was always ever a lot of places I've been, we always were the last ones that brought in because they're afraid you're going to say no. The truth is most of the time the reason we say no is because we got blindsided with something that we had no preparation and no context around before you went out and decided to purchase it. And then you come to us and go, hey, I bought this brand new whiz bake thing. It's going to do all this cool stuff for us. And it was like, no, because you're giving it access to sensitive data and it's going to take that sensitive data and put it out. It can't ever have access to this data because it's privacy data or something like that. And it's like, didn't you guys ask these questions? And then you see the deer in the headlight look and number 1A, either all they go is well, you guys are just trying to keep us from being able to make revenue and causing problems or they're just completely shocked. The fact that they, if they even have a little bit of an inkling of understanding what risk is that they open the company up to all this risk. Because then you're like, look, I don't have enough information to sign off on this, but if you are going to force me to, you're going to sign off on the fact that you understand that this is risk very bad and you're taking responsibility for putting it into an environment that typically changes the conversation. Right. When, when you make them be responsible for it, it typically changes the conversation.
42
00:37:51.000 --> 00:38:04.000
Well, and that's where this is. AI is not doing anything different. It's just making it faster. Like these are the same problems we've been having. All we've done is shine a magnifying glass on it and make it. It's like, you know when you get
43
00:38:04.000 --> 00:38:06.000
put a supercharger on it. Yeah.
44
00:38:06.000 --> 00:38:50.000
Like I was about to say, like you, you, you. When your kid learns to drive and they turn 16, they get their driver's there. It's the same risk they're gonna wreck their car, they're gonna be a shitty driver. If you put them in a 1991 Land Cruiser that is slow as just they have more time to react. Whereas if you put them in a freaking Ferrari day one, they're just gonna get into trouble faster than, than they. It's not the car's fault, it's their fault of being a shitty driver. It's just we've basically AI has, has put the supercharger on. You put a supercharger, you added nitrous, you put slicks on it and then you put an unex. Designing the process behind the driver's seat. And then you're surprised when it goes haywire and you run into a brick wall.
45
00:38:50.000 --> 00:40:45.000
Yeah, yeah. I mean, and, and I think you, you talked about it a minute ago and like I was talking about the fundamentals, the things that we have struggled to do properly over the years. And it's, it's not from a lack of understanding. I think a lot of times I think it's a lack of, it's a lack of time, it's a lack of resources. Because nobody wants to go in and completely redesign an identity platform. Nobody wants to take time to do that because it's not something that happens overnight. And the problem is you wind up piecemealing it over so many years you now have wound up with this convoluted mess that again, over privileged, over entitled, too much access. People have got permanent that have got their accounts running as service accounts, all this stuff going on. And now you're trying to untangle that mess and go back and figure out what really needs to work. And then you're always going to break something when you messing with identity access. Right. The developers wrote a program in a way that has to have local administrative access. And it's like why, why did you do that right. In today's day and age? But they may have started with an old program that they just never updated that or instead of plugging into an ip, an idp, they decide to write their own identity into their tool. Right. And now you got this whole isolated identity thing and it, it's always been ugly. Identity is, is one of those things that, you know, if, if magic wand, enough people, enough money, enough time, you would go, we're just gonna, we've, you know, looking at a couple companies I've been that did massive amounts of M and A over the years and you'd go back and go at one point I was at, we had 120 ads. Active directors. Had 120 different active directors.
46
00:40:45.000 --> 00:40:46.000
Yeah.
47
00:40:46.000 --> 00:41:00.000
Trying to get them all to work together. And it's like, you know, wouldn't it just be easier just start over new, just to build a greenfield ad and migrate everybody into it?
48
00:41:00.000 --> 00:41:06.000
Yep, it would be better, it'd be cleaner, be safer, be more secure. There's a Lot more work that goes into it and a lot more work
49
00:41:06.000 --> 00:42:41.000
that goes into it. A lot of. But there. And, and there's going to be some disruptions and stuff like that, but ultimately you have many. You'll have many less exceptions than you will if you just do it now. There's some new. And this is where AI and some of the new startups are coming out. I've seen some really cool stuff around that identity space and helping you fix that problem that are coming on, that are hitting the market and looking out there. There's some really cool stuff that's I think, going to help us with that. I've seen a bunch of people throwing around the term or throwing around. It's like shadow AI is here and it's going to. But it's not. It's a reality. I hate tell everybody it's been a
reality for longer than the last year. You know, shadow AI has been around for a while because you may think you're blocking some of it. But what you realize is that the tool. Take Microsoft Teams, for example. I was, we were doing teams meetings and we started getting notes from an AI tool that we had never heard of, right? And these were some relatively sensitive meetings. And it's like, hey, who authorized you guys to use that, that AI tool to take notes? So it was like, all right, we'll go block that website. Well, it still was happening. And then we found out, well, it's because it's available in Microsoft's app store inside teams, so you can just download it and install it. It's like, you know, you've got to remove your. And that's where I go back to. You have way more AI agents than you think you do and more AI tools because everything's got it built into it now. It's just like in the, like I said, in the old days of API, everything had an API built into it. And if you didn't know how to turn it off and protect it, people could connect to it. Well.
50
00:42:41.000 --> 00:43:33.000
And it also goes back to, you know, least privilege, the foundational type things. But it really comes down to we have to be able to go into. So an API is a great example, right? So we've got these APIs and I see this way too frequently, and I know you do as well. Like, I don't know what I'm setting this thing up. I don't know what access I need. I don't truly understand the API or what the thing that I'm asking for needs, what access it needs, what it's trying to do. So I just give it all. It's like in any, any rule on a firewall as I'm testing out this process. But then you never go back and actually remove the things that it doesn't actually need. It's just like Active Directory. Like Active Directory. By default they want to have all of these capabilities and a firewall rule. If I'm having to do authentication through a firewall, I remember setting up firewall rules where, you know, the default is, is they want 3,000 ports open to be able to authenticate to Active Directory. Yeah, of course that makes it easy. But you can also restrict it down to like five.
51
00:43:33.000 --> 00:43:34.000
Yeah.
52
00:43:33.000 --> 00:44:38.000
That it absolutely needs. And the rest are just, just, yeah, nice to have. Or you can reconfigure them or turn them off. Wouldn't you rather have five holes instead of 3,000? But most people just go with the easy button and say, well, PALO has an automatic, you know, I'm not calling out palo. It's all firewalls, right. It's, it's the easy path to do those things, but we've got to get to that least privilege. And it really comes down to understanding your business process and documenting, getting, you know, you look at nuclear power plants, right? They truly are air gapped. They are because they're using data diodes and they don't have any Internet connection whatsoever in or out. And the only thing that does goes through a diode. So it's physics. I can't reconfigure it. I can't. There's no vulnerability on it. AI is not going to change that. Right. So if you get to your foundations, then all of these things aren't as big of a problem. It's basically just going to force us to do all the things that we as practitioners have been screaming forever. Do the foundations really well. Understand your process. Don't have 120 active directories as an AI tools. Let's simplify the stack so we can actually understand what the things are. And then we have less coverage and less vulnerabilities over our heads.
53
00:44:38.000 --> 00:48:06.000
100%. Right. And one thing I, you know, talking about just access, right. Like you brought up great things. You, you highlighted what I think is the biggest problem with an SDLC at most companies is they've got a dev environment that has almost no controls in it. Then man, you're right. They can just wide open do everything they do so they don't have to fix it in depth. Right. Then they might move to a test environment. Although they should be moving to a test environment. That test environment should be a hundred percent security control wise and path network. All the controls all the way the servers are configured. All that should configuration should all be identical to production. Never is. Nope, never is. So because as soon as it moves from dev and it breaks in test they go in and instead of fixing the problem, they reconfigure test to allow it to pass and then when it goes into production, it breaks and then you wind up having, you normally wind up having to put an exception in and then track. And then over time they'll get back to fixing this. Right. Or maybe, maybe, maybe. Right. And well that's the other dirty secret about audits. Right. You make anybody pass an audit of any kind. Yeah. You just turn it off and break the system while the auditors are there. And then when the auditors leave, you go back and flip it back to the position it was and go oh well, we'll fix it before they come the next day time. And then you know, just, it's, it's a perpetuating cycle of never slowing down enough to actually fix the root problem, which is not doing the basics right. It's not doing the basics well. And you should be, you should be doing this stuff through all your environments. Now every IT operations folks, and I've been one of those folks before, business leader is going to tell you you're slowing us down, you're costing us money. Money. Right. They're go, that's, that's what they're going to tell you. And yeah, I am. But you know what, you're also costing us money because it's breaking and we're having to go back and do all this other work or maybe we're having to buy other tools or buy more licenses to be able to people put more fences up to protect the crap that could have been fixed in development because like I said, because the developer decided they needed local admin access to do what they're doing instead of, of doing it a different way.
Right. Or you know, so that's a very important. And then the other thing I was going to bring up is I was thinking about this the other day is like, you know, every company I've ever been at, everybody's got way more entitlements than they should in their, in their access. And most of the time they never know about it. You can most of the time can turn it off. You can look at it and go, well they've never used this. Susie has never accessed this server. She shouldn't have had access to it anyway. She's never even logged into it to. Let's just take it away. Well, the problem is, is Susie's not going unless she's all of a sudden an insider or is curious. That happens as well. It's probably never going to be a problem. You write an AI agent that you tell it to go do a job efficiently, it will search through all those nested entitlements and all that stuff and try to figure out how it can gain access to this stuff because you've asked it to do something efficiently. So it's smart enough to go look for that access and to figure it out. Like the new, like we said with the new OpenAI and new mythos stuff, they were designed to go find vulnerabilities and they're very, very damn good at it.
54
00:48:06.000 --> 00:48:07.000
Yep.
55
00:48:07.000 --> 00:49:35.000
And the same thing with identity stuff and the other big piece that this is brought up, since we're talking a little bit about development, all this is the build versus buy conversation is back on the table. A lot, a lot of people are like, should we just build this? And I, I don't disagree with that. Where my only push back a little bit on building in house is something you. We've kind of talked about with the going back to the basics is, is our team have the bandwidth, they got the knowledge, they have the skills. Do they have the bandwidth to document this and to keep it up to date and to improve it over time and not get married to what they built it once and set it and forget it if you're not having that discussion. When you're comparing the cost of, of buying a tool and the. Like you said, these smart people are building these tools that do. Yeah, I can write data discovery. I can write an agent to go do data discovery within 15 minutes. Right. Depending on my operating systems and my storage environment I'm in. And it'll, it'll come back and do all that for me and I don't have to buy another tool to do it, but am I going to keep it up to date? Am I going to be able to design three of them for every different cloud environment I need to talk about? Or do I just buy one that has all the connectors that connects into my sim, connects into my vulnerability management, connects into all this stuff and know I'm going to get updated versions of it and not have to tax my team? My internal.
56
00:49:35.000 --> 00:49:38.000
That's what, that's what you're paying for, right? That's what you're paying for.
57
00:49:38.000 --> 00:49:39.000
Right.
58
00:49:39.000 --> 00:51:48.000
You're paying for that. And I've gone through this conversation over the years and it gets back to, if you think about it, you know, the days of the mainframe, the days of these in house built products that are really great, but what happens when that engineer retires? Now, AI can help you with some of that, figuring out what it's there. But again, it gets back to if you build it, then you have to support it, which means you need to be testing it, you need to be pen testing, you need to be, you know, vol, like all of those types of things. So it's more than just, yeah, I can save a hundred grand because I don't have to go buy this tool. But it's going to end up costing you more than that in the long run. And if you're not a software development company, why are you adding additional capabilities that you don't necessarily have in house and you're not, not, you're not really good at. It's not that you can't do it. You've got smart people to your point. It's like, do you really want them? It's like, I don't mow my yard. It's not because I can't. I've mow my yard for years. I did it as a kid. I'm very capable to do the things. I don't do it because I don't. That's not where I should be spending my time. Like, I don't configure firewalls for clients. Like, I'm a consultant. I've configured thousands of firewalls in my career. But you don't want me doing that. Like, of course I can, but it's not the best use of my time. So that's where we need to really be thinking about this. Just because you can build it cheap and easy with AI up front doesn't mean you should. And what is the risks that you're bringing on? And it's not just a cyber risk. It's. It's really a bigger risk to your operations, to your understanding. Recovery, disaster recovery. What happens when you build this process and it's part of your overall process and then you don't know how it works. And then I leave because I built it. And it's not necessarily because it's a cyber vulnerability risk, it's a vulnerability to your operations and availability. Like, that's the piece that we don't always think about AI. It's not just cyber, it's also the supportability, the functionality. And then when you have 16 different things designing and you have no standards, then you've got this hobble podged, you know, massive scripts and things like that, that, yeah, it works as long as you stand on one foot and hold your head over your hand over your head. But what happens when one of those things break the whole freaking cards, all the dominoes fall and now you're sitting here standing with all the pieces
59
00:51:48.000 --> 00:51:49.000
and
60
00:51:48.000 --> 00:51:50.000
you don't truly understand how it all worked together.
61
00:51:50.000 --> 00:55:14.000
Well, you just brought, you just brought it back to the cyber, to the resil business resiliency problem, right? You get used now or relying on this tool to do a business process and if the person who built it or it's, it's evolved to a point that it is critical to the
business, what happens when it stops working? What happens when it breaks? What happens when it gives the wrong data output? It. Right. How does that. Were you prepared to get something back up and running when that, that you now have your business relying on it, right? If you're not thinking about that as you're building this stuff out, you're in for, you're, you're in for a world of hurt, right? Because you're going to go down one day and not be able to get back up in a reasonable amount of time. And that's, that's one of the big things again, having the right people in the room when you're discussing this stuff and not let anybody be having an AI conversation in a vacuum, right? We, we all can get in our own vacuum and echo chamber and think that we're coming up with the most incredible thing. I was talking to a guy the other day and he's like, yeah, when cowork first come out, he said my wife was asking me what I was doing at like 2 o' clock in the morning in the office. He said, I'm in here solving the world's problems, right? He said because it's so cool what this thing can do. I've done figured out how to solve all the world's problems. But then you look at it and go, what happens if that problem changes, right? Or the degradation of the model goes, starts, starts giving me wrong information, right? So it's one of those things. It's like everybody gets real excited about it, but make sure you get the right people in the room. And you got to have somebody in there with a couple of voices of reason and it's normally going to be somebody from the security team. There's probably going to be somebody from the IT operations team and go, how do we support this? And you know how much this is going to cost to do what you're. Because you're talking about this much cycles in the data center or in your cloud environment. It's like this is a lot of money. So you gotta, you've gotta have those realistic conversations. And some of this is reminds me very much of the early days of shifting to the cloud. Right. Is because when you had all the on prem data centers and people were like, oh, it's gonna be so much better because we can move so much faster, we can scale so much quicker in the cloud. 100%. Absolutely. Absolutely. But have you looked at the budgetary cost of what we pay for to run power today, what we pay for our Internet circuits, what we're paying for all that stuff? All right, now we have ingress and egress cost. Now we have storage costs that we kind of had a set figure for. It didn't burst up on us all the time. We've got these consumption models. We now have a latency because we didn't understand. You don't understand what an application turn is on, how many times it's going back and forth, forth. So you got to have all these conversations people when early on it's like holy crap, it's way more expensive to run this stuff in the cloud than it is to have it in an on prem data center. Yes. Are there some efficiencies to it? And there's still
very good reasons to do it in the cloud, but you really gotta have somebody who can ask those questions go, how many tokens is this going to take? What is our ingress and egress cost going to be on data moving back and forth between these. How are we going to protect it from going out and connecting? They're putting data through a public API. You got to have everybody in the room having that conversation to lay that plan out and document it. And again, back to the basics. It's a build it as a program. Don't do it in a vacuum that somebody's got the greatest idea on the planet because it's going to cause more problems than it is going to fix.
62
00:55:14.000 --> 00:55:24.000
Yep. A hundred percent. All right, man. So we, man, we talked about a lot of doom and gloom, but also some exciting things like what? What's the thing that you see that you're most excited about with all this coming out?
63
00:55:24.000 --> 00:59:58.000
So I think it's going to move us into the world that I've been preaching for the last five years. Even before the AI became, became real big was because we could still do it with machine learning and with APIs is I think this moves us into what I see as continuous the government or Fisma and FedRamp NIST. I'll call it continuous monitoring, continuous configuration management, call it continuous governance, call it continuous compliance, call it whatever you want. But I think we are now going to see the ability to use AI and APIs that are out there to continuously pen test our environments, to continuously measure drift of configuration, to continuously measure how our environment is configured across all of our compliance goals. Because now I don't have to sit down, think about a global and I used to do this, the global company where I have multiple network teams, firewall teams. We'll go back to our firewall discussion. I can sit down with the guy in the US and ask how the firewalls are configured there and he'll tell me how they're configured. The ones he knows about about. I go to Europe, I got the same problem. He's going to tell me how his firewalls are configured. The ones he knows about. I go to apac, I get a different. Now what I figured out from having this conversation is A none of them know how all the firewalls are configured across the globe and B they don't even know about how all the firewalls in the region are configured because each one sub region inside that or each office may have somebody who's done something different. And I've spoken spent hours now of multiple people's time asking questions. I no longer need to do that. I can have the tool connect to the API, pull the configuration, measure the configuration against a known framework and tell me how I'm configured. And I can go, okay, why is this one different? I can tie it into my change management, understand how changes are happening so I can control drift much better. I can really change the way we do governance and compliance and audits. I should be
able to show a screen to an auditor and go here's how everything's figured configured, here's the report and how do we know that we're doing the right thing? Because I have all this data that I'm able to near real time or in real time interrogate and do that Same with the pen testing. Right? The pen testing is going to be I can now using AI agents and the tooling out there, I can pretty much pen test continuously in today's regulatory and compliance world, I'm still going to have to pay somebody to do an external one because you can't pen test yourself for most of the frameworks, right? You got to do at least one that's an outside external, but you can do enough pen testing that anytime a change goes into your environment, a major software change, a major hardware change, a major network change, any of that process change, you can now test your environment and see, did I make my environment it better or did I make it worse? And hopefully you can get to this predictive side of it and go, hey, we've got a development environment or a sandbox environment built so close to production. I can run this through with an AI agent and run it through and go, where is it going to break? What's going to. And this is going to change vulnerability management too, because I think you can. You should be able to say, if I roll this patch out for let's call it Java, if I roll a Java patch out in our custom developed application, what's going to break? That's always the biggest pain point with development teams, right, Is answering that question. We don't know if it's going to break it. So we got to spend all this time researching and all that. Hopefully the AI stuff is going to allow you to do that in a sandbox and go, all right, if you change this library or you update this thing, this is the problem you're going to have in your application. You need to do these configuration changes along with this to allow that to smoothly rol into production and get out of the world. Where we roll something into production, it breaks in production and then we have to roll it back out and then go figure out and backlog it and do all this. It's all this gyration that eats up everybody's time. I think that is the most exciting stuff I see from a security and a risk standpoint. We're going to be able to know before we roll out in a lot of cases, whether or not we're breaking something or we're opening ourselves up to a known risk. Now the zero day thing gets a little more tricky. But I think think seeing what Mythos and OpenAI's product is doing, there's going to be some neat things coming from that too. Now, depending on how that gets rolled out, I kind of hope they don't just give it to everybody. I do like the fact that they're keeping it pretty locked down to the. To the providers right now. I have. That's one of the.
64
00:59:58.000 --> 01:00:02.000
Yeah, but that's just theirs. That doesn't mean that other ones don't have it.
65
01:00:02.000 --> 01:00:05.000
That's true. These
are just the big boys. And I guarantee you other ones have it right.
66
01:00:05.000 --> 01:00:08.000
Those are just the public facing ones. So that doesn't mean.
67
01:00:08.000 --> 01:00:39.000
Those are just, those are just the big boys. Right. That are, that are trying to do most. Yeah. Trying to do mostly the right thing. I won't ever accuse any of them of being 100% altruistic, but. Yeah, but exactly. But that's what's exciting to me is I think it will give us a much, much better understanding of how our programs are operating and how, how they're drifting. Because it's always a problem. Because you can build the most perfect cyber security program day one by day
68
01:00:39.000 --> 01:00:43.000
1.5 minutes later, once you turn it on.
69
01:00:43.000 --> 01:00:47.000
Yeah, it's changed. Somebody went, ah, that didn't work. Turn that off. Flip that switch.
70
01:00:47.000 --> 01:00:48.000
Yeah, yeah.
71
01:00:48.000 --> 01:01:46.000
Or somebody forgot to flip the switch on that one device. So I think, I think that's one of the things that I'm super excited about is that ability to move into that continuous monitoring. And then when you have to answer the question to a CEO of what keeps him up at night or what should he be worried about? Are. You know, the famous one is you always get asked as a CISO is are we secure? And you always have to answer, well, for what we know about today, yes, you can now give them real information and do better budget planning and better strategic planning, resource planning and go, hey, we have this gap because we see it, we need to address it or we're good here because I can interrogate the system and go, hey, this new vulnerability just came out. Out. We just heard about it for this new CVE is out. Where does it exist in my environment and how long is it going to take us to patch it? That question should be almost instantaneously be able to answer going forward with these tools that are out there.
72
01:01:46.000 --> 01:01:48.000
Eventually we'll get there.
73
01:01:48.000 --> 01:02:05.000
Yeah, eventually we'll get there. But I mean that's what's exciting to me is we're going to see that dramatic shift and then the one thing we'll, we'll just throw this out there as another topic for another day. Day. This conversation becomes much different and much more important when Quantum gets more affordable. Huh?
74
01:02:05.000 --> 01:02:09.000
Yeah, that, that changes everything. Also that, that, that unlocks doors and
75
01:02:09.000 --> 01:02:14.000
all the things these two together become. Yeah, scary, scary.
76
01:02:14.000 --> 01:02:15.000
But they could be good as well.
77
01:02:15.000 --> 01:02:38.000
Right. So I think it's going to become very good. But it's also very scary because we're so unprepared for what it's doing now as an industry we add that superpower of, of being able to, to compute at quantum levels. It's, it is incredibly scary because bad actors are going to have that probably more readily available before the general populace does.
78
01:02:38.000 --> 01:02:45.000
Yep. Well, it's called Action Man. How do people find you? What's going on in your world?
79
01:02:45.000 --> 01:04:08.000
Easiest way to get a hold of me right now is on LinkedIn. You know, I'm, I'm, I'm in between roles right now, looking for my next thing, but doing a little consulting like here on the side, helping some companies out as we talk about, talk about these problems. Always trying to have conversations with, with the CISO networks that are out there across the country that I'm involved with. Because guys, we can, we can, we need to bring these problems up and we need to be voices of reason and, but we also need to be voices, when I say voices of reason. We also need to be willing to listen to people and go how do we address, address the problems and help them be more efficient and be more resilient. So let's get out there, learn if you're not using AI in your own personal sandbox. Absolutely. Get out there and mentor other people in your community as well that are not in our industry. Get out there and talk to the kids that are coming out, Explain to them what this stuff is. I mean unfortunately I just saw and this is a great example, this kid here in Georgia just got 19 years old, just got a serving prison for hacking. And it's because he, he taught himself how to do it. But he went, he got onto the bad side. Started out as video game cheats and then figured out that he could dump out more data. Right. And it, got addicted to it. He said it was a bigger adrenaline rush than driving 120 miles an hour.
80
01:04:08.000 --> 01:04:09.000
Yeah.
81
01:04:08.000 --> 01:06:03.000
Look for those kids who've got that capability, get them focused to be on the, on the good guys side, you know, help they, the talents out there. Make sure you're also, and this is my other last call to action. Make sure you're having conversations with your HR people and explaining to them that they should not have ridiculous requirements on technology length that hasn't been in existence for that long. Degree programs. Look for aptitude, attitude, drive to be part of this. Be realistic with what you're asking for people to do. Be willing to train people in your environment because if you bring someone in, in and are willing to train them and teach them and get them built, they will stay around a whole lot longer than the transient environment we're in today in the job market where people are job hopping from job to job looking for the big paycheck or non toxic leadership. That's the other thing people leave about, right? But find this talent, get at the, the job. The talent pool exists, we're just not allowing it to have access to us. Now granted, I do know there's a problem with the AI getting so many job resumes, but talk to the local people in your community. You'll find some really cool talent out there. Just have a conversation with them. Get out to your local ishackers, your local, your local hacker cons, your local conferences that come through and put yourself out there and spend some time talking to these people coming up in our industry and explain to them what they're going to be dealing with and explain to them not everybody wants to be a ciso. Not everybody. It's a, it's a, it's, it's a job that, you know, everybody I know that's not doing CISO work now calls themselves a recovering ciso. It's funny, it's become part of the vernacular that we all do because we've all lived through so much of this stress. But get out there and, and share your knowledge with everybody. That's why I do what I do. Because sharing knowledge is the best way to get our, to move our industry forward. Forward.
82
01:06:03.000 --> 01:06:17.000
Yeah, man. Well, thanks for your time, man. Good conversation, man. There's so much coming up on the horizon and I appreciate your insight and definitely reach out to Ken and all the things that he does and helps and we all need to be in this battle. So thanks again, Ken. I appreciate it.
83
01:06:17.000 --> 01:06:20.000
Thank you. Eric, great to see you.
84
01:06:20.000 --> 01:06:44.000
Thanks for joining us on PrOTect IT All, where we explore the crossroads of IT and OT cyber security. Remember to subscribe wherever you get your podcasts to stay ahead in this ever evolving field. Until next time.