In this episode of CYBR.SEC.CAST, Michael Farnum and Sam Van Ryder sit down with Alfred Huger, co-founder and Chief Product Officer at Command Zero, to explore the rapidly evolving world of agentic security operations. They discuss what organizations should consider before moving toward an AI-driven SOC, from establishing clear performance baselines and understanding costs to giving AI the business context it needs to make better decisions. Alfred also breaks down the importance of role-based access controls, governance, and balancing human oversight with the increasing speed of AI-powered attacks. The conversation offers a practical look at what CISOs should be asking as agentic SOC technology moves from emerging concept to real-world deployment. This episode is sponsored by Command Zero Things Mentioned: Command Zero - https://www.commandzero.ai The Agentic SOC: Why Security Operations Must Reimagine Itself - And Fast: https://www.commandzero.ai/resources/agentic-soc.pdf CYBR.SEC.CON Discounted Hotel Rooms and Parking - https://www.cybrseccon.com/resources Careers Fundraisers - https://www.cybrseccareers.org/fundraiserevents Do you have a question for the hosts? Reach out to us at media@cscgroupllc.com In this episode: Host: Michael Farnum Host: Sam Van Ryder Guest: Alfred Huger Director: Bill Brenner Producer: Lauren Andrus Music by: August Honey Keep up with our Conferences and Events: LinkedIn X Facebook Instagram TikTok CYBR.SEC.CON. OT.SEC.CON. CYBR.HAK.CON. Keep up with CYBR.SEC.Media: LinkedIn X Facebook Instagram TikTok YouTube Learn About CYBR.SEC.Careers Non-Profit Efforts CYBR.SEC.Careers Subscribe to the podcast: Apple Spotify YouTube Listen to our other shows: CYBR.HAK.CAST CYBR.Minded CYBR.Signal Thank you to our Media Partners: CYBR.SEC.CON. Breaking Through in Cybersecurity Marketing CYBR.SEC.CON. and OT.SEC.CON. OGGN (Oil & Gas Global Network) UtilSec Packet Pushers CYBR.SEC.CON. and CYBR.HAK.CON. BarCode Podcast Cyber Distortion Podcast Kill Chain Radio The Phillip Wylie Show Vulnerable U
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
SPEAKER_03And welcome everybody to another episode of the Cyberset Cast, the CybersetCon podcast. As usual, here with the man with the nice head of hair, Sam Venre. We have got a very timely conversation today with the chief product officer over at Command Zero and also co-founder at Command Zero, Alfred Huger. Yes, it involves AI, but it also involves security operations centers and agenci SOC and the how you get there.
SPEAKER_03What's the best path? He's got a really good ebook that's put together that we'll include in the show notes. Great conversation, I think.
SPEAKER_02It's time lamp, as you say, and it's super complex and something we could literally do a gazillion episodes on, given the fact how fast this stuff is moving and how much we've already experienced in such a short period of time. So very interesting.
SPEAKER_03Yep. It's a great conversation. Let's jump over and talk to Alfred. Let's go. Hey Alfred, how you doing, man? Um well, how are you, Michael? Doing well. Great to have you here, all the way from Command Zero, all the way from the Great White North up there.
SPEAKER_01How's the weather? It's six a day, it's quite beautiful. It's been pretty smoky, I think, like most parts of the United States. Well, yeah. Today's pretty clean, so it's nice. That's good.
SPEAKER_03Yeah, y'all been sending a lot of smoke our way, I think.
SPEAKER_01Yeah, hey, I got mine from Oregon. That's where mine's first. So we're getting some from you guys too. Fair trade.
SPEAKER_03We're about to get some. I think at least the Caribbean's about to get some dust from Sahara, but that's a good thing because that dust keeps the uh the hurricanes away. I'm okay with that. Interesting.
unknownOkay.
SPEAKER_03That's yeah, it's an interesting weather pattern we get here during the summer, still in Houston and the Texas. I want to before we get moving and go into talking about CybersetCon, do want to divulge that Command Zero. I want to say thank you actually, Command Zero, for sponsoring this podcast. As usual, these conversations are going to be all about what's the good stuff, right? We're not, we don't do product pitches and all that. That being said, Command Zero will absolutely have the opportunity at the end of this to talk about what they do.
SPEAKER_03And Alfred, you being the chief product officer. I think you're probably the perfect person to do that.
SPEAKER_01I hope so. We'll see.
SPEAKER_03Yeah, we'll find out. If not, we can just edit the whole thing out. But no, we'll keep it at. All right, Sam, we need to go talk about CybersetCon and everything else we've got going on. So let's jump over there.
SPEAKER_02All right, everyone. So we have some announcements. Let's roll through these real quick for you. So obviously, we've got the big conference, CybersetCon, coming up on the 15th and 16th. That's in Houston, Texas, at the George R. Brown Convention Center downtown Houston. Tickets are on sale. Um, the full agenda is out. Go check out the agenda. It's absolutely fantastic. And so many great speakers, so much great content. You will not gonna want to miss it. Plus, we have two new villages that are joining us. We also have some specialized tracks here, uh, one around OT, one around hack, so more uh technical in nature, exec, and then AI.
SPEAKER_02So those, and then we have lightning talks uh on the energy stage, um, a broadcast alley. So we have that for podcasting, et cetera. And then we've got uh the launch pad uh competition, which will be uh as well, which is really cool show showcasing new technologies uh coming to the market. So and we'll have more.
SPEAKER_03Yep, so much going on. It's crazy. Um away from the fun stuff. Sorry, but we do have to talk about some updates. So one thing if you have not got your hotel room as of this recording, um, the Hilton has expanded. That's when's directly connected. It's extended expanded their discounted rates to September 4th as well, or extended. And that room after that is going to go up over$100 a night if you wait. So go get that because the inventory of those are very low.
SPEAKER_03Um, another urgent update, if you don't know about this, the Georgia R. Brown's got a lot of construction going on. They're expanding. So over the next two, three, five years, I know it's like a lot going on there. Uh, they're expanding the Georgia R. Brown convention center. Uh, one of the garages that we thought was going to be open, we found out is not going to be open. So uh you can park either uh at the north garage, which is directly attached, or you can park at the tundra garage, which is a little bit of a walk away from the Georgia Brown.
SPEAKER_03So go make sure you those aren't the only places to park around there. Uh go out to cybersetcon.com slash resources. Let me say that again. Cybersetcon.com slash resources. If you scroll down, you'll see where all those are located. You can find out where to park. What's next, Sam?
SPEAKER_02Well, let's talk a little bit about cybersec media. So we have four podcasts now. So that's cybersecast, which we're doing right now, right? We have cyberhackcast, uh, we've got cyber signal and cyberminded. So you can check all those out on cyberstechmedia.com again, cyberwithout the e. And you can also subscribe to those on Apple, Spotify, YouTube, or wherever else you get your podcasts.
SPEAKER_03Yep. And if you want to get all of it in the central place, cybersecmedia.com is where you'll find all of our conference talks. You'll find all of those podcasts, all that as well. Um, it takes us about a month after CybersecCon, after any of our events, anyway, to get all of the talks up. Um, we've got about 170 speakers or something crazy like that this year. Um, so if you haven't already, go sign up uh to be a member, get in the newsletter on cybersecmedia.com because we'll push all of those out in our newsletters when those come out.
SPEAKER_03You don't want to miss that. And that's completely free. And we do not sell, we do not give away your info. That's really for us just to communicate to you.
SPEAKER_02Yep. And so last thing, we've got uh cybersec careers, which is an incredibly important part of what we do, right? So we're working uh with them through a managed partnership agreement, and it's we're and they're hosting two fundraisers that we're part of, right? So the first is uh partnership with uh with CruiseCon. So if you use our code Careers15 at checkout, you'll get 15% off your booking. And a portion of your purchase will go towards our programs, which is super great, and you can feel good about going on a cruise. We're also hosting a casino night event on November 5th at St.
SPEAKER_02Arnold's Brewery, just north of downtown Houston. And you can find those tickets on cybersteck careers.orgslash fundraiser events, and we'll have that linked in the show notes as well.
SPEAKER_03And all 100% of that profit goes to support our programs, and that's mentorship, that's the Taylor Austin Broussard Memorial Scholarship, uh, that's our youth seccon, everything we do there. Um, the applications for the scholarship, by the way, are uh going to be open on September 15th. Uh, and it also goes to the user group that we put on as well. So we've got youth seccon, we've got user group, we've got mentorship, we've got the um the memorial scholarship, Taylor Austin Bruce Ard Memorial Scholarship.
SPEAKER_03Uh, go to cybersecareers.org, also without an e, cyberseccareers.org. You can learn more about that, what we're doing. And by the way, if you go there and you can't be a part of any of those things for some reason, you can just give a donation. This is all to help students and veterans get into cyber and get them better, uh, better along, farther along with their their careers and help them get into cyber and find out what it's all about. So it's a great, great program to support. So please just consider a donation there.
SPEAKER_03We would really appreciate that. All right, before we keep going, I'm gonna give a quick word about today's sponsor, Command Zero. So the average security operations center faces more than 11,000 alerts every day. And over 67% of SecOps teams report severe alert fatigue. I mean, it's just a real thing. Meanwhile, it takes more than 200 days on average to detect a breach. I've said this before many times, and I'll say it again: adding more tools usually means adding more alerts to an already backed up queue.
SPEAKER_03And that means that fatigue we just talked about with your SecOps, folks, it just gets worse. What security teams need is a platform that can identify real threats and actually finish those investigations so that fatigue goes down. Command Zero investigates breaches, they investigate phishing, identity threats, malware, insider activity, just about every other kind of threat facing your organization. It's an autonomous AI-assisted SOC platform that handles tier one all the way to tier three investigations at AI speed, all while making sure that those humans are still in control of it.
SPEAKER_03Command Zero is not going to rely on rigid playbooks. Instead, it queries data wherever it lives, applying the expertise of a seasoned security analyst. It's really important to your organization. And last of all, we all know transparency in SecOps and AI SOC is a big conversation these days. That's why every step Command Zero takes is documented and made visible, auditable, and reproducible. And all of that I just talked about, well, it doesn't have to take forever to get it going.
SPEAKER_03Command Zero can be up and running in under an hour and really start delivering answers in seconds. All of that because Command Zero finishes investigations. All right, let's get to the rest of the podcast. All right, we're done with CybersecCon, all the updates. Got the commercial break. Thank you again for Command Zero for sponsoring this podcast. Alfred, all of our guests get at least 10 minutes. You can do less, you can you can't do more, but you can do less if you want to talk about what got you into this crazy world of cybersecurity.
SPEAKER_01Oh man, I certainly won't take 10 minutes, but I had a full head of hair when I started this, so it was a while ago. So you have to bear with me while my memory banks kind of get plumbed. I started working as a system administrator for a subsidiary Nintendo in the mid-90s. Yeah, it was a lot of fun. Actually, it was pretty cool. And then while I was in that role, protecting what we were building was really important. Plus, security was really interesting. So I started spending a lot of time learning how to secure systems on IREX. If anybody remembers that, it's it's been a while.
SPEAKER_01And then I the founders of that company were really well off, so I asked them to invest in a company and I went and started another kind of cybersecurity company in 1996 along with a gentleman named Oliver Friedrichs. And we ended up building vulnerability management software and scanners and sold it to McAfee, which was, I guess, is still around. It was called Cybercop Scanner. I don't know what they renamed it after it was acquired, but while we were there, we built a whole bunch of other stuff. And then he and I went and formed another cybersecurity company after that, along with a bunch of other folks called Security Focus, which had bug track and um we built an early warning system from OR outbreaks and a bunch of other stuff, and then ended up getting acquired by Symantec, where I got to work for a long time, which was a really cool place while it was still a place.
SPEAKER_01No longer is, that's not my fault. And and that we got really heavily involved in the endpoint experience then, as you might imagine, because that's that was jam at the time. They did a lot of stuff, but endpoint was what they did most of. So we went and did a cloud-based EDR company that was then acquired by a company called Sourcefire, which does Snort. Of course, we all mo many of us remember Snort. And it was it went into Cisco, and I spent probably uh gosh, like eight or nine years at Cisco. That was my last adult job before jumping back into startup land. And there I ran engineering full for Snort and for next gen firewall and email security and a bunch of other stuff, which was a pretty cool experience.
SPEAKER_01Big, very big company, probably the largest one I'd ever worked at the time. But yeah, so that's been like the last 30 years compressed into a very small time frame. Yeah, it's been a lot of fun. I love this space as much as I did the first day I got into it. And if anything, it's like it's even more insane and interesting now, which is great. And now I'm the chief product officer over at Command Zero, which is another startup that I've helped co-found. And it's in the writ large, it's in the SOC space, but it's in it's a Gentec SOC to be a little more specific for the time and place. Yeah. So that's me.
SPEAKER_03Yeah, I was having that conversation the other day about the um how it feels like the I don't know, it's not the mystery, but the excitement feels like it's back these days with the AI, with the entry of AI and everything. It's like everything else before was kind of like, all right, well, another ransomware attack, you know, and now it's like, oh, this is like really interesting stuff.
SPEAKER_01There was there was like a lot of variation on theme, you know. You you take a little step up in the same product category, maybe something a little more interesting than Zero Trust or EDR on the firewall. But they these are like now we're seeing like logarithmics, sort of like steps up in what we can do that we couldn't do, not just us. Obviously, it's always been asymmetric, but now the attacker has access to neuromancer. So it's like William Gibson got the light. Yeah, I I heard that, which is actually pretty cool. I want to go see that. Yeah, um yeah, so it's it's a it's gonna be a frightening brave new world that we're moving into for sure.
SPEAKER_01And I don't think that's hyperbole, I think that's just common sense.
SPEAKER_03Yeah, no, it's real, it's absolutely real. It's scary and exciting, and but at least it, you know, I don't know, keeps us from being bored.
SPEAKER_01Yeah, maybe we'll see if the attack the advantage can move to the defender for change with this. And if it's like this is the only time in our history that I can think of where this might be the place where we can see that take place. Because I've always felt like we've been on the back foot for the last like the 30 years that I've been in the space. It is you know it's unfortunate we need to do what we do, but there's never been a time where I could sit down and be like, look, I can credibly tell you that your security posture is going to be significantly stronger than your attacker's capability to get through it. And may maybe that changes now.
SPEAKER_01We'll see. I think at least now the doors potentially could be open. It'll be interesting.
SPEAKER_03Well that I think that's a good segue to talk about our main topic today, which is which I like. This is you call it an ebook? Is it an ebook?
SPEAKER_01It started off as a one-page document that that our that our marketing team got a hold of and became an ebook. But yeah, so I think it qualifies as an ebook. It's a bit like that. Yeah, it's a little more than one page now.
SPEAKER_03But yeah, but it's still pretty concise. I mean, it's not like you know, it's 50 pages or what is it, like 16 pages?
SPEAKER_01Yeah, we tried to we tried to keep it to the to to its salient points. It's a it's it's a topic that hasn't had a lot of discussion before because I think until pretty recently socks as a product offering to customers haven't really been a thing per se. There's been platforms that serve that purpose. There's been a lot of different products that you tie together to build a sock yourself and and whatnot. As a space, I think a lot of people that we talk to are getting their heads wrapped around the fact that they can, in fact, bring something like that in-house. And then you introduce the idea that it's agentic in nature, which is a completely different kettle of fish that you have to address as well.
SPEAKER_01Like you're getting beyond why I would buy that. I think most people are pretty pretty comfortable understanding why they want it. That's why they're talking to somebody like an AI sock vendor. But like what differentiates them? What am I looking for? What are the what should I be concerned about? What are the metrics I should be adhering to? All that stuff is still pretty fresh. It's a it's maybe not a brand new category, but it's pretty close to a brand new category. Usually you used to get this from an MDR and MSSP, which was closed software that you didn't know.
SPEAKER_03Yeah, that's what I was thinking. Like yeah. They the the EDR goes to the MDR where but a lot of the MDR companies still looked at it being mostly managing endpoint security versus the whole thing. Then that category got expanded in the XDR to some degree where you were able to see a lot of different telemetry, but it was still you were still running into the issue of like how do I handle all this telemetry? Who's really taking care of it? And then that's when managed XDR, MXDR, or MMDR, MEDR, whatever you want to call it, all the M, yeah, all the acronyms drove me crazy.
SPEAKER_03But I had several customers that wanted that managed side of it, and they came in different flavors. They came in flavors of I've never run a SOC before, so I need somebody else to run it. Or, you know, I built a small kingdom and did a good job putting the people in there, but we can't keep up with all the alerts. Everything's so new, we just can't do it. We don't want to do 24 by seven. We we expanded into an international company, so we need this follow the sun model, but it is okay. Now you got to go to an MSP to do that stuff.
SPEAKER_03And it just it ended up being a quagmire every single time, it feels like.
SPEAKER_01Yeah, those are the same. I think those those sort of profiles are still pretty similar. You get we also see a lot of um in-house socks that believe that and they tend to be in larger or or better funded organizations. They tend to believe that with the advent of agentic technologies writ large, that they can tackle a lot of the things they had their sort of more junior analysts or their L1 analysts have that done with agenic workflows and then sort of reposition, hopefully reposition those team members into tasks that they were never able to get to before, that they think need better investment, et cetera.
SPEAKER_01And that's really interesting to see happen. Because that whole tier one space has been sort of where it's a pretty thankless job, right? And it's thankless jobs get very expensive very quickly, right?
SPEAKER_02Yeah, that's the other piece of it. I think there's a real opportunity to actually support the business by doing exactly what you say, have those people do something that's more effective for them than the menial tasks that AI can take over.
SPEAKER_01That's right. And I think it also I think that was a tough one to argue. Maybe a year ago I would have been a little skeptical about it in some cases, but uh the the technology's come so far in such a short period of time. I think it's also from a risk reduction perspective, much more compelling for CISOs and for organizations that have to be able to quantify what their capabilities truly are. You think about the size of like the shape of a SOC has always been weird to me in the sense that I think it's a zero-sum game. If it moves past tier one, its next stage is you're calling Mandian, right?
SPEAKER_01And your tier one analysts are always your most variable skilled, lowest paid team members, right? That's where you start. You cut your teeth there. But your tier three and two and above, your most experienced team members are in the back of the room playing catch for whatever gets past tier one. And it seems to me that you'd want those people to be better engaged. Those people that have the experience are who you want with the hands on the levers. And I think that the promise of an Agenic SOC is that is you can credibly handle the blocking and tackling and actually put more of those tier one assets into tier two and three and push them all towards the front of the problem, or or dealing with things like bone management that you might have sucked at before or whatever it might be.
SPEAKER_01And I think that's credible. I don't think that there were technologies that could credibly offer that a year ago, a year and a half ago.
SPEAKER_03I that's when they started coming out. It was probably a solid year ago at I think it was at Black Hat a year ago. Because that's just that was just what last as a time of recording, just a couple of weeks ago. And I think it was the first time I ever saw 7 AI. And that was at Black Hat a year ago. And then I saw another one. And like I knew this was coming because I had I did some work with some of those old school MDR slash socks that some of my customer customers had that they were like, look, to Sam's point, like it was so expensive to keep level one analysts in there and to make them at all effective because they were had this massive growth because nobody wanted to deal with it, but a Gentex sock wasn't ready yet.
SPEAKER_03So they just kept throwing college kids in there to cover the load, and they kept screwing things up. They couldn't follow basic processes. It's nothing against them, they're just not mature in working in that world. Doesn't mean they're dumb, they just didn't know how to do it. So I had a customer basically like, look, I'm gonna get off of this MSSP because they've pissed me off a million times. I'm gonna go to this one in the interim. Literally signed a one-year contract because he was like, in a year from now, I think agentic sock will be viable and I can start moving.
SPEAKER_03And they they just signed up for one and agentic sock, and they're making headway, and it like it it was a night and day difference from what I saw just six months ago, how they were able to handle it.
SPEAKER_02Yeah, so I built for a living and I've seen that. Sorry, Sam. No, please. I I guess my question to you is what's the path? If aside from the story that Michael says, what do you see typically where a company needs to migrate or has the opportunity, what fundamental things do they put in place before they actually deploy something like this?
SPEAKER_01That is the I think the exact right question that the Little CISOs don't ask themselves before they have the conversation about moving to an AI sock. And it's really important. I think first off, laying out your objectives, what it is that you want to achieve and making them measurable. And it sounds like common sense, right? You go into any IT project like that, but that doesn't happen. And I think that you have to understand what it is in your SOC environment today, it is you're looking at changing and have an actual realistic baseline for it. That could be easy to do around cost. It costs me this much to you know provide this much risk avoidance and staff my team.
SPEAKER_01It could be around things that are that are related to cost but are more around posture, like how many cases are you able to address, how many things go to escalation, whatever that might be. Have them understood before you go to an AI sock vendor. And also understand that if you can, to the degree you can, understand the quality of your current work. How good are you at what you do? Because you won't be able to measure how well your your AI sock vendor does for you if you don't truly know what your existing baselines are. Same thing goes for your MDR or your MSSP, because that tends to be, Michael, as you mentioned. A very common path forward is I'm unhappy with my MDR MSSP.
SPEAKER_01Often the underwritten tone there is I don't like how much I pay them as well. And I will say that it is an unproven statement that AISOC will end up being cheaper than MDR and MSSPs over time. So look, just that's the elephant in the room. Vendors are eating cost right now because customer acquisition is more important than profitability. And you have to be aware of that, right?
SPEAKER_03That's any AI company right now that's sitting running tokens, right? They're just they're like a screw it. We've got a lost leader. We've just got to do it.
SPEAKER_01It costs money. It costs money. And that's that's the thing I always but that yeah, you bring up a great point.
SPEAKER_02It's across the board BTEC, right? It's when you go in, you your acquisition costs to get and this is where customers get caught off guard is at a renewal date where things become more mature, we have a better understanding. They get a bill and like, oh, hold on a sec. This isn't what I originally signed up for. And because yes, because you were an early adopter, you came into it early, you know that you just understand. I would say from my perspective, when I talk to customers, just understand that this is going to go up potentially, or it could come down, depends on how the type meets meets or gets developed, right?
SPEAKER_02But those are a concern anybody should have.
SPEAKER_01But from a but from a purchaser's perspective, I mean, so baselining is really important, but also protecting your ability to maintain that investment in the future is important. So the conversation has to be had with the vendor that you're choosing. What is my price in control? How can I guarantee that if I fall in love with you, that next year I don't have to mortgage my family to keep the product? And a lot of one-year deals are super common right now because people are shocked at how fast things are moving. They want to see how things pan out with different vendors, and that's appropriate. Um, multi-year deals obviously lock you in after some advantages you don't see in a one-year deal.
SPEAKER_01But if you do a multi-year, make sure that the pricing is cogent and that you can afford it in year two. And there are ways to negotiate that. And your vendor should be able to give you clear, transparent models on how they do their pricing and a commitment for at least year two and three on their increase in pricing. Otherwise, you're gonna go. I don't think so. Okay, anymore.
SPEAKER_01I but I'll explain why I say that. I think any AI-driven system, or yeah, I think a multi-year deal is always gonna give you a better experience, period. As you integrate it more deeply into your current processes and workflows, you deal with muscle memory, all that stuff. The ability to consume business context and make good decisions out of it right now in at least some of the AI vendor sock platforms is really powerful. It's fast. It doesn't, it's not supervised learning. Um, so it doesn't require the same volume of data or the same amount of retraining or any of those things. So if you extract the right data about your environment and make sure it's kept current, don't get stuck in the CMDB nightmare of once write, never write again.
SPEAKER_01Um I think that you could actually credibly move AI sock vendors out after a year without as much pain as you had before. And that's why there's 65 vendors in this space right now, because it is a technology that that can strap into an environment and work pretty quickly if it's done right. And of course, of the 65, there's probably a handful that do it right.
SPEAKER_03But that I mean that that's where I get stuck on this, Alfred, because I if you're saying just very focused on the SOC portion of it where you're not doing like essentially level one. Like everything you do on level one is I see an alert come across. That alert is not generated by anybody, but you know, some of your tools that are in the environment. It's not generated by the agentic sock, it's generated by EDR or whatever.
SPEAKER_03You know, you've got detections in your SIM or whatever that is, that feeds into the level one analyst. That level one analyst goes, okay, this looks credible. I'm gonna pull all this evidence to figure out whether it's really credible. And then I'm gonna hand if it is, I'm gonna hand this off to somebody to go take care of this because I, or maybe I'm gonna, it's slow enough to where I can click a button and isolate, and then we can figure out whatever that is. So agentic sock is doing that now. What I've seen that's not as easy to plug in, and what I haven't found in a lot of agentic socks be is like the feedback mechanism that's actually going into the detections, like the feedback mechanism, or not the feedback mechanism, but the contextual mechanism that's bringing intelligence in.
SPEAKER_03So because the sim ingestion from a you mentioned cost perspective, the sim ingestion and those kind of things are what the where the extreme cost is in a lot of places. How do we get that fixed? And that's not as easy to plug in because that's those are harder, that's a harder mechanism. That's a harder thing to do. Does that make sense?
SPEAKER_01Yeah, it makes complete sense. And AI saw it deployed without appropriate context, what you're referring to, the ability to extract context around that alert material. That is what separates out the wheat from the chaff in terms of an offering that works well. So that is the ability to understand ground truth about assets just as a basic example, like who owns it, who should be accessing it, who typically accesses it. Has it been involved in false positives before? Has it involved in red teaming exercises before? What are the what's the purpose of the place where it's deployed? Is it in your shop floor in Singapore or is it a finance system in Manet?
SPEAKER_01Who are your users? Who are your identities? What blast radius do they have? What are their job functions? What are notes about previous cases? What files tend to lead to FPs in your EDR? Is it do you have a bunch of scripts in production that you build cars around that your EDR goes wild about every time it sees it? Those are things that an analyst knows because they've been in the seat for a year or two. And they intrinsically are able to absorb that information and quickly decide the narrative around a case and whether it's something that needs to get escalated, closed, false positive, true positive, whatever that might be. A good AI-driven SOC extracts that information and uses it to navigate a case.
SPEAKER_01Otherwise, it will navigate a case and try to make answers based off the data it has. And often, even alone, that's better than many tier one analysts that you'll encounter. But it's not going to be as good as your good ones. Because the good ones have time in the chair, they have intuition, they have human curiosity, which an AI still doesn't have. And if you want consistency, which the SOC needs to be successful, business context is 100% a key function in that. And so a system that understands what context matters, how to weigh it, how to how to analyze it in the face of other data and conflicting hypothesis data.
SPEAKER_01You know, so like I may find that Michael does travel for a living as a salesperson, so him logging in from Singapore might be okay, but there may be other data that conflicts with that. So I can't use that context as a free skate to avoid alerts for Michael, just as a basic example. They get a lot more complex. And I think what separates out a good AI sock from something that was vibe coded over a weekend is that level of nuance and control. And I think one of the things that is really important and and but not sexy is RBAC. We've all read about what happened to Hugging Face and all the other agents gone wild.
SPEAKER_01An AI SOC gives agents, if not architected correctly, direct access to everything you care about and direct access, all your O365, all your email, all your administrative systems for identity, all your identities. You have to have a system that has world-based access control, it's extremely thoughtful, and does not give agents direct access to your systems. Because we can't guarantee that they won't slip the reins of control. If anthropic and open AI can't, I guarantee you a security vendor can't. So the architecture needs to support the agent simply not interacting with them.
SPEAKER_01It needs to move through a set of intermediaries or some architecture that says it doesn't speak directly to your crown jewels. Yeah. And so our back matters as well.
SPEAKER_03I think that's gonna be the if anything slows us down from a defender perspective, is getting us potentially getting way too cautious with the guardrails and not being able to keep up because the there are no guardrails on the attacker side. I think that's gonna be an interesting race in the future.
SPEAKER_01And I think that that is the rogue agent thing reading Bob's email that he shouldn't, that's a real concern. But I think more of a concern is an entitled agent that has good R back, making bad decisions and containing things they shouldn't. Yeah. I think that's a more practical use case. And when we first started doing AI sock, most people wanted human in the loop, meaning I want to see the case before I remediate. Yeah. But when you see attackers leveraging open weight models and frontier models, I think the luxury of you allowing for human in the loop is time limited.
SPEAKER_01Yeah. So you're going to have to accept some business risk there at the end of the day, which but but you have to also say this platform has the relevant controls in it that I at least believe should work most of the time. But you're going to end up making mistakes like you do with any technology. But the question is whether the platform you have is going to help you make as few of them as humanly possible and correct the ones that you had. But yeah, I don't think you'll be able to wait for human in the loop. I think that's going to be very quickly a thing of the past. Because in if not this calendar year early into next, it it's always been an inefficient model.
SPEAKER_01I think now it's going to be insufficient.
SPEAKER_03There's one of your uh, I won't name them on here, obviously, but there are there is one of your competitors that very early on just said, screw it. We're not even to talk about human where they just went fullborn. I was like, Y'all might be a little too early on this, but okay.
SPEAKER_01There's a comfort level with that. And there's good good risk and governance reasons you don't do that. And some places will never be able to do that. They just won't. They're just not allowed to do that, no matter what. But I mean, like with everything that we've all experienced in the last forever, the attacker is forcing the narrative in this case. So I don't know how long it is before most attacks are driven by large language models, but I think that is an outcome that we're looking at. Yeah.
SPEAKER_03Well, Alfred, fantastic conversation. It's a great resource. We'll include this in our show notes. Um, definitely not overdone. I read through this, it's really good stuff. No, thank you. I appreciate that. Appreciate the time you put into it and providing a resource like this. You're sponsoring this, bad boy, so give us a good 30-minute pitch. Or not 30 minutes, sorry, 30 seconds.
SPEAKER_01Yeah. At the end of the day, Command Zero, our focus is on building an agentic SOC that provides both L1 through L3 use cases. We think L1 is a must-have, it's a must-be this tall to ride feature in a SOC, AI SOC, but ultimately a lot of effort needs to be put into cases that move beyond that. It's a tier two and three, incident response, human-led investigations, identity-led, all those things. So our platform tries to be able to provide that level of capability. And hopefully, for the customers that are on board with us over the time, we'll move into other functions as well outside of just the SOC.
SPEAKER_01But first, we need to crush that one out. And that's where we are today. Awesome.
SPEAKER_03Well, and I want to also thank you for sponsoring CybersetCon. Of course. And are you coming out to speak?
SPEAKER_01No, I'm not because that falls in the middle of my holidays. But we have a lot of excited people that are. So awesome.
SPEAKER_03We appreciate y'all doing that. That's gonna be fun to have you on there.
SPEAKER_01That'd be awesome.
SPEAKER_03All right, Sam, any parting thoughts on your side?
SPEAKER_02No, this is a topic we could probably talk for, another one that we could probably talk for hours. And I didn't even touch on the OT side of things here, which I won't, so we'll keep it short. Sweet. Oh gosh, this gives good stuff. Yeah. Good stuff.
SPEAKER_01Thank you very much, gentlemen, for having me today. I love having an opportunity to talk to other people in the industry. So this is always a lot of fun for me. Usually I'm heads down in front of a computer working on building a platform. So it's nice to not have to do that. So thanks for the break. You ever want to come out of the cave? Just give us a call. We're I will we'll take your call. Thanks, gentlemen. Have a great week.
SPEAKER_03You too. Thanks everybody for joining us on CybersecCast. And remember, uh, community yields better resilience. We'll see you on the next.
SPEAKER_00This has been a cybersec media production. Cybersecast is hosted by Michael Farnum and Sam Van Ryder. It's directed by Bill Brenner, produced by Lauren Andrus. Our music is by August Honey. The views and opinions expressed in this show are those of the speakers and do not necessarily reflect the views or positions of any entities they represent. This show is for informational purposes only and does not render or offer to render personalized advice. Subscribe now so you never miss an episode.
SPEAKER_00You can find all our podcasts, articles, blogs, and conference talks on cyberstechmedia.com. That's cyberwithout the e. And follow cyberstechmedia on LinkedIn, X, Instagram, Facebook, TikTok, and YouTube. You can keep up with our conferences by following us on LinkedIn, X, Instagram, and Facebook at Cyberstech Events. And you can learn more about our events or buy tickets at cyberstec.community slash cyber dash sec events.
In this episode of CYBR.SEC.CAST, Michael Farnum and Sam Van Ryder sit down with Michael Santarcangelo, founder of ValueSmith, to discuss delivering value that others can actually recognize. With three decades in the...
In this episode of CYBR.SEC.CAST, sponsored by FireMon, Michael Farnum and Sam Van Ryder sit down with Jody Brazil, founder and CEO of FireMon, to explore how AI is reshaping network security and why traditional...
In this episode of CYBR.SEC.CAST, sponsored by Endor Labs, hosts Michael Farnum and Sam Van Ryder sit down with Peyton Kennedy, Senior Security Researcher at Endor Labs, to explore the hidden risks of AI adoption,...
About this episode: On this week’s episode of CYBR.SEC.CAST, Michael Farnum and Sam Van Ryder sit down with Bob Ackerman, Managing Director at DataTribe, to share lessons from decades of building and investing in...
About this episode: In this episode of CYBR.SEC.CAST, Michael Farnum and Sam Van Ryder sit down with Eve Security founder and CEO Nadav Cornberg to explore the challenges of securing agentic AI. They discuss AI...