Matt: Today on cloud security today, we have something a little bit different for you. I have on researcher Regav Singh, whose work explores the human side of organizational security. Now, his work draws on dozens of CISO interviews, and I wanted to have him on to discuss something that's been talked about a lot in the industry, but oftentimes it's kind of just tangential, no research behind it. That is the topic of CISO burnout. Now, Ragov in his research, as you will hear, he goes beyond the headlines. He talks about why burnout goes beyond long hours, why reporting to the CEO doesn't solve all the problems or guarantee real authority, and very apropos, how AI could trade alert fatigue for explanation fatigue. If you love what you're hearing on cloud security today, I would love for you to pause the episode right now and give us a five star review wherever you listen. to your podcasts. It helps a lot with Reach. Hope you enjoy the episode.
Matthew Chiodi: Rakav, welcome to the show.
Raghav: Thanks, Matt.
Matthew Chiodi: So I wanted to have you on the show because you've done some really interesting research around cybersecurity. You you took an angle that not a lot of people have taken. So you go into many different areas, but you've researched CISO burnout, CISO CIO power dynamics, AI and security. And now, or at least most recently, as far as I can tell, you've dug into how attacks begin before. execution. Now those are some seemingly very diverse topics. In your view, like what connects these topics and what led you to focus on the human in organizational side of cybersecurity?
Raghav: Yeah, so I actually landed into cybersecurity research completely by accident just because I I attended a couple of lectures from you know, a professor who taught cybersecurity and assurance. And that really, you know, opened the doors for me into getting into cybersecurity research. And when it comes to CISOs themselves, so I had a CISO friend of mine. So we once started talking and at one point of time he mentioned that he was just burnt out and considering exiting the role. altogether. So till that point of time, I always thought that being burnt out and being stressed are the are one and the same things. But after talking to him and a bit of research, I learned that I learned how different they are. So for example, when you're stressed out, you feel like there's too much to do. But when you're burnt out, you feel like you have nothing left to give so that just put things into perspective for me. that there's just so much risk when the person who is responsible for guarding your organization is burnt out. you know, the risk just being financial, personal, or even reputational, right? so that's how I landed into working on the CISO burnout project. you know, burnout simply isn't just about somebody working too many hours. because I would hear things like, I'm respe I I'm responsible for this risk, but I cannot really control the budget. Or things like, you know, I told them about the problem six months ago, but nobody really acted upon it. and a lot of CISOs also talked about, you know, how security is always blamed for slowing the business down and sort of, you know, talked about as being you know an innovation inhibitor of sorts. So all of these things, all of these you know, quotes from CISOs, that really just, you know, took me naturally into questions about power. Like you know, what authority does the CISO all actually have? And how does the relationship with the CIO actually work? Who controls the resources, who gets access to all the power and the board, and who can who can actually have the liberty to say no? Right.
Matthew Chiodi: Mm-hmm.
Raghav: And Then AI came into the being and AI made all of those questions even more interesting because now you have CISOs who are being asked to secure something that the organization simultaneously wants them to enable very quickly.
Matthew Chiodi: Hmm.
Raghav: So so that that was another one. and as for the attack conception work, that really just takes the same instinct and applies it to the attacker side. So normally you know, security becomes very interested once. Somebody starts doing something which is recognizable as malicious. But what we are doing in the attack conception research is that we are asking what happened before that. So for me, the common thread is really just the idea of conditions rather than the events themselves. because if you look at a breach, a breach is just an event. burnout itself can also look like an event. And the same can be said for you know the conflict between the CISO and the CIO, but usually there's a process underneath it. and that's probably what has kept me interested in the human and organizational side of cybersecurity. so of course the technology is being selected and it matters a lot, obviously. but when the technology is you know being selected, it's being funded. And even being interpreted by different people inside the organizations, right? So a technically excellent security control inside the organization with poor communication and
Matthew Chiodi: Mm-hmm.
Raghav: you know, maybe misaligned incentives as well, all of that can still produce a very bad security outcome. So I don't really see cybersecurity as purely a technology problem. I see it more like a socio technical problem.
Matthew Chiodi: Mm. I appreciate how you're approaching this methodically. And I will definitely put a link in the show notes to to your research. I think our audience will be very interested in kind of your methodology for some of these things. I know you just said that you spoke to a number of CISOs. I'm sure people are going to want to dig in and say, you know, how open were they? Like how did you find these people? We don't necessarily have to go in that now, but I think that'll be really useful. I think these conversations and this information. Is going to be really useful. So let's let's double click into your burnout research. So in the burnout research, you identified six categories of stressors. But your more important conclusion, at least from what I took away for it, it seems to be that they interact and reinforce one another. And I think you alluded to this just a couple seconds ago. Can you explain the difference between contingent, complementary, and substitutive stressors using a real world CISO example, anonymized obviously.
Raghav: Yeah, sure. So yeah, I think that was one of the more important pieces of that study. so again, it was an academic research study. So initially, even though we were thinking about just identifying what are the stresses that contribute to burnout, one thing that we really wanted to study is how do those stresses interact, and if they do at all, right? so yeah, we identified those six broader categories of stresses, but I wouldn't want somebody to just walk away from the paper thinking that okay, here's this checklist of six things that make CISO stressed, right? so the more interesting part in at least in my mind was that these stresses interact. So a simple example could be that imagine if you're a CISO and you don't have enough resources. That's obviously stressful. but now put that CISO into an organization that also has a weak security culture. So suddenly that resource problem becomes much more worse because every request that you have to make for the resources requires you to first convince somebody that cybersecurity is worth spending money on.
Matthew Chiodi: Mm-hmm.
Raghav: so the effect of that one stressor depends upon the presence of another. So that's what we mean by contingent relationship. right. And the other one was compliment complementary stressors. so This is where, you know, two problems actually reinforce each other. So for example, if your security team is understaffed, right, so because if you're if you're understaffed, the CISO then starts getting you know, he he starts he or she starts getting dragged into operational work. So
Matthew Chiodi: Mm.
Raghav: now the CISO has less time for being strategic in their communication with senior executives. And because that's communication then suffers, so the executives may not really fully understand why the additional resources are required. So the resource problem creates a communication problem. And the communication problem makes the resource problem even more harder to solve. Right.
Matthew Chiodi: Mm.
Raghav: So that's how you get into a loop. so that was the complementary stresses. And then the third one is substitutive stresses, which I find found really interesting. so in substitutive stresses, you might assume that you know, as a CISO becomes more experienced, their stress should just decline. And you know, some stresses absolutely do decline. So you bec you become better at handling incidents, you know how to talk to the executives, you even I mean, you you've seen some more things before, but
Matthew Chiodi: Right.
Raghav: then there are different pressures that appear. So now you're presenting to the board, and you know, there may be some regulatory exposure. So your decisions are more visible, you're more involved in the enterprise politics, there's potentially personal accountability associated with the decisions you make as well. So one form of stress gets replaced by another one.
Matthew Chiodi: Mm-hmm.
Raghav: so that was one of the things that really changed how I thought about burnout. So you know, it's not necessarily just a pile of independent stressors where you just you know remove one thing and everything gets better. So we found that it's more like a system. So if you change one part of the system, something else can change with it. So that's what we are trying to get at when we said that okay, certain stressors you know, interact with each other.
Matthew Chiodi: Are, you know, between these three, the contingent, complementary, and substitutive, is there, did you and you know, I didn't see this necessarily in the research, but is there a greater weighting on any one of these more than the other? Like are one of them heavier, more stressful, or is it individual? What is that what is that relationship look like with those? We're able to quantify that.
Raghav: So so interesting thing is, so once we had developed this model, we actually showed it to a couple of CISOs just to just for their validation and thoughts. And one of the CISOs, I won't name them, but so they made an interesting comment that the weightage of each of the stressor categories, they would differ depending upon which CISO you talk to. Because there could be, you know, different nuances like, you know, which industry do they hail from? What is their reporting hierarchy? What is their you know, organizational security culture? So I think it would differ from person to person. but what was really interesting was that all the three all all the you know different types of stressor categories, they really just emerge from the organization security culture. So that
Matthew Chiodi: Mm.
Raghav: was really, you know, just just the initiating force, if you will.
Matthew Chiodi: You mentioned seniority when we were talking about substitutive stressors. In your research, did you find any correlations with seniority? You know, if this is someone's first time time being a a CISO versus their second or third, was there any correlations there? Did you guys dig into that? I'm always curious about that. Cause you're right. You would think, you would assume that if someone's done this two, three or more times, that there would be certain stressors. that they're like, I've seen this before. I've done this two or three times. Or maybe it was before I was even a CISO I dealt with this because I led whatever it was, right? I led the company's efforts against, you know, for GDPR, et cetera. What did you find there?
Raghav: Right. So we didn't really directly find something there, but I don't know if this answers your question directly. But one of the things that we are working on right now, which sort of relates to this, is you know, what is the relation between the maturity of a CISO
Matthew Chiodi: Mm.
Raghav: and how they manage burnout. So based on that, we actually like you know, in the burnout study, we interviewed 37 CISOs from different levels in different industries. So we were actually able to slot each of those 37 CISOs into three categories, CISOs one CISO 1.0, 2.0, and 3.0. And I'll just quickly explain the difference between each.
Matthew Chiodi: Please.
Raghav: So CISO 1.0 is somebody who is only technical to the extent that you know, they they are only technical to the extent that they are not able to communicate well with the board as to what they require in terms of resources. So imagine them going to the boardroom and throwing all this technical jargon when the board actually just wants everything in terms of dollars and cents, right? So they're not able to communicate well with the with the board. CISO 2.0 is also fairly technical, but they have a risk-based sense. So they understand risk. But even they have difficulty in communicating their needs with the board. So even they struggle. CISO 3.0 is somebody you know, who who You would see as a business enabler. So they are very strategic. They have a say in which direction the business is moving. And this actually reminds me of a couple of CISOs who mentioned that, okay, when you look at the C-suite, there's actually a capital C in a small C. So in the capital C, you have people like CEO, to an extent, the CIO and the CTO, and you know, the others, but unfortunately the CISO falls under the smaller C. So even though they are a part of the C-suite, they don't really enjoy the kind of autonomy that the others do. Right. But for CISO 3.0, they are the ones who would fit under the, you know, the capital C because they are business enablers. So that is a research that we're still, you know, working on right now. and you know, we are looking at what are sort of the type of stresses that you know these three types of CISOs experience. What are the what are the different nuances? What are some of the coping mechanisms that they talk about? So so for example, quite a few, and again, I don't want to generalize, but just based on our very small sample of 37 CISOs, some of the C some of the CISOs in CISOs 1.0, they actually talked about negative coping mechanisms
Matthew Chiodi: Mm.
Raghav: in order to address or manage their stress. but when we are talking to people who we had slotted in CISO 3.0, they were talking about positive coping mechanisms.
Matthew Chiodi: Mm.
Raghav: right? like, you know, building up their you know, interteam relationships and building relationships with the board and s and whatnot. so yeah, I I don't know if that directly answers your question, but that's one place that we are heading to with this research.
Matthew Chiodi: I love that that concept of those three archetypes. I think that's really powerful. And I'm curious, and again, maybe this is part of the updated research, the research that you're working on now, but with that CISO 3.0, is that really more of a is that a look at just the individual or is it a combination of the individual and their program? Because if it's the individual and the program, I'm curious, you know, and I we'll we're gonna talk a little bit more about security culture and whatnot. But I'm just curious, like when you're looking at those three archetypes, how much are you looking at the individual themselves versus the individual themselves and their current organization?
Raghav: so right now we are only looking at the individual. to be very honest, we I mean I would concede that, you know, we hadn't taught he we hadn't thought about, you know, bringing the program in.
Matthew Chiodi: Mm.
Raghav: we were thinking about you know that being some sort of like moderator relationship. And you know, but yeah, I guess I learned something today that, you know, we
Matthew Chiodi: Yeah.
Raghav: should we should actually bring the program in. yeah, I think that would be a cool idea. Like I said, we are still in the initial stages of this research.
Matthew Chiodi: Sure.
Raghav: But yeah, that would be one thing that we'd love to consider. We should consider. Yeah.
Matthew Chiodi: Yeah, I think the the my brain immediately goes to that simply because I I was wondering, you know, if you took someone who you would classify as a three-o and you put them into an organization that had a weak security culture, was you know, every CISA would say that most CISAs would say they're under-resourced, but let's say they were severely under-resourced. If you took that three-o, put them in that org, would they change to The two to like how much is it their individual? And I'm not, you don't have to answer this question. This is the kind of the thing I always wonder about. Or maybe someone who's a 3.0 wouldn't accept a job in that type of organization because of their level of self-awareness that they would ask, they would know these questions. And you mentioned the interview that I did with Rinky a couple months back. One of the things she talks about is how she interviews the organizations before she takes roles for that express reason. So something just to maybe think about as you kind of r as you redo this research. So if I can be
Raghav: I agree. Yeah. That that's actually pretty good food for thought for me. Yeah.
Matthew Chiodi: I love that. Well, let's let's you know, culture itself I is extremely important, obviously in any organization, any function, but your research specifically points to weak security culture as a root amplifier. you talk about expectation asymmetry and how CISOs communicate risk and a maturity paradox where seniority changes stress rather than eliminating of it. So Which of those would you say is the most misunderstood by CEOs in in boards?
Raghav: so I would say that the maturity paradox, simply because it's a little counterintuitive as well. so we tend to assume that somebody who has been a CISO for 15 to 20 years should find the job easier than somebody doing it for the first time. and in some ways they do that, right? Because they have developed some coping mechanism mechanisms, they understand the politics. They've even probably dealt with some major incidents before. So they know how to communicate the risk. But then seniority also changes the kind of stress that you experience. so because you know, you're no longer only worried about whether the firewall is configured correctly or whether the SOC has enough people.
Matthew Chiodi: Yeah.
Raghav: now you're actually talking directly to the board. So you're dealing with regulators, you're you know. Dealing with the board members, and are a part of the strategic decisions as well. And your name may be attached to statements about risk as well, right? and also you know, people are increasingly expecting you to provide certainty aro around the things that are inherently uncertain as well. So the experience doesn't necessarily make stress disappear. Sometimes it actually reconfigures the stress.
Matthew Chiodi: Mm.
Raghav: and It just reminds me of one of the CISOs that I interviewed for you know, like a like another study. And at the time that I had recruited him for the for the study, he was still a CISO. But when we actually met for the for the for the interview itself, he had taken up a director of security position in another organization, which was in his mind sort of a demotion.
Matthew Chiodi: Mm. Sounds like it.
Raghav: And he was completely happy with it because he said that at least I don't have to take that responsibility and that stress. And mind you, this was somebody who had close to 20 years in the in the field, and maybe I think 10 or 11 years as a CISO. And he actually told me that a lot of the CISOs that he knew, they would happily, happily take a pay cut or even a sort of a demotion of roles if it meant having more freedom to work on. And lesser accountability as well. so so yeah, you know, the the maturity paradox was one. the other one that I think that executives underestimate in some terms is the expectation asymmetry. because the the CISO may be looking at a very messy technical reality which has a lot of uncertainty in it. And then somebody in leadership suddenly you know. Suddenly but understandably asks a completely reasonable question that are we secure? Right? and it's such a broad question
Matthew Chiodi: yeah.
Raghav: to to answer, right? So so what does that mean? I mean the real answer might involve talking about threat models, probabilities, dependencies, vulnerabilities, who are your third parties, all kinds of things. But then you only have maybe three or at a maximum five minutes with the board. So somebody has to absorb that complexity and translate it into something actionable without making it so simple that it becomes misleading. Right. and very often that burden often sits with the CISO.
Matthew Chiodi: Do you feel that, you know, obviously I've been in the cybersecurity industry for now 26 years. And so I'm, you know, part of this maybe group think to a certain perspective. I think there's a a thought in in the cyber industry that CISOs are unique in this. And I don't know if they are or not, but in these things you talk about, the expectation asymmetry, the maturity paradox. And I don't know if you maybe I don't know if you've looked, but outside of cyber, does this also exist in other C suite roles or is cyber truly different? And if you don't have the research, that's fine. I'd love just your thoughts on that.
Raghav: I mean, based on the research, okay, and when I say research, I mean purely academic research.
Matthew Chiodi: Sure.
Raghav: so based on purely academic research, no, we haven't seen that. So we have seen, you know, CISO to be a very new C suite role. And a lot of the stresses and a lot of the complexity that they have to go through, they haven't been experienced by the rest of the C suite. So in that terms I would say that, you know, it's like what they're experiencing is really new. Yeah.
Matthew Chiodi: And I think that makes sense too. Even if you just literally look at the evolution of the titles, the only one that would even be close to that would be the CIO, right? The CIO is a role that's been around now for, you know well, I guess the nineties would have sometime in the mid to late nineties where the CISO title only really showed up in any type of weight would have been, you know, single digit two thousands, right? So you're looking at a a decade or more difference. So I could see that. I could see that. So the the takeaway right here for all our CISOs that are listening is that you are a snowflake. So some of the challenges that you're dealing with really are unique. And so it's good that we are we're talking about this. So one one other follow up question on this kind of whole topic of security culture before we we move on to some of your other research is are there observable warning signs that would tell a CEO or a COO or whoever that that this is becoming an organizational security risk rather than merely an employee wellness problem, which is often how it's when we're talking about pure burnout, which you kind of talked about some of those differences there. Other observable warning signs?
Raghav: I mean, yeah, there are. so for CEOs specifically, I mean, I would look for organizational symptoms rather than just asking whether the CISO seems tired. I mean sure.
Matthew Chiodi: Might not get an you might not get an honest answer there.
Raghav: Right, right. I mean, everybody has their pride. I mean, they they might want to seem very macho and not admit to it. or they might not have enough FaceTime. Who knows? But yeah, I would look at you know organizational symptoms really. so for example. is security reportedly being brought into the major initiatives after the important decisions have been already made? Right? And is the CISO having the same sort of resource conversation every quarter? questions like you know, are the risks being escalated but not really actually being resolved? Is there high turnover in the security organization? Another thing like, you know, checking if the CISO is spending most of the time just firefighting rather than actually shrinking, actually thinking strategically. right. And I think the bigger question for not just the CEO, but the board members as well and the senior executives as well is that are the senior executives executives learning about risks for the first time when something goes wrong, right?
Matthew Chiodi: Mm-hmm.
Raghav: Because if That's the case, then at that point it's not just an employee wellness issue. you know, the organization's ability to communicate, to reorganize, to recognize and even respond to the cyber risk, that's starting to deteriorate. And that's a security problem. so if you systematically exhaust the person that your organization is relying on to challenge. some risky decisions, you know, it's just an eventuality that burnout will become a governance problem.
Matthew Chiodi: Well, I think in where my head goes to immediately after that is reporting structure, right? Since you know, you you mentioned that in the C suite there's kind of the big C's like the COO, the CEO, CFO. There's also but then there's this kind of the smaller C's, which the CSO is oftentimes part of that. So one of the things we've been debating for many years in cybersecurity is where The CISO reports, right? Should they report to the CIO? Should they be a direct report to the CEO, etc.? I've been to some organizations where the CISO reported in to chief legal officer, right? So so tell us like, does your research suggest that the reporting line itself is the challenge, or is it a proxy for more important issues, which we've been talking about, right? Board access, budget authority, escalation rides, independence.
Raghav: Yeah. So so take whatever I'm going to say with, you know, like a grain of salt right now, because I sort of am a critic, not just of academy, not just of academic research, but but of the industry as well. so I think that the industry sometimes turns this into too much of a yes or a no question. Should the CISO
Matthew Chiodi: Okay.
Raghav: report to the CIO? my answer is that the reporting line matters, but I wouldn't stop there. Because what does that reporting line actually represent? That's the question, right? I mean you have to answer questions like, okay, can the CISO get to the board when necessary? Can are are they able to escalate a concern, even if the concern involves a decision which is being made by their own boss? Do they have autonomy? Do they have budget authority? are they being involved you know early enough in major technology decisions? you know, just to be able to meaningfully influence those decisions, right? So things like you know, those questions are much more revealing than just making it a yes or no question, like should the CISO report to the CIO? because I mean you you could put the CISO directly under the CEO on an organization chart, and you could still give them very little actual power. That can happen. You know, and maybe engineering sits somewhere somewhere else, infrastructures sits somewhere else, and the budget sits somewhere else altogether. So, you know, the business units then just make their own technology decisions and you then have something that I would describe as you know, you you still have that formal independence, but then you still have operational dependence as well. But then the opposite is also possible that the CISO could report to the CIO and have a very healthy relationship. In fact, the second CISO paper that we are currently working on, which is about just the power dynamics between CISOs and CIOs, we were initially looking at dyads. so when I say dyads, I mean you know, CISOs and CIOs hailing from the same organization. And we were interviewing them in such a way that neither of them knew that they were being interviewed because we did not want
Matthew Chiodi: sneaky.
Raghav: Yeah, because we did not want to get into a situation where you know, if you're talking to a CISO, they would recommend us talking to their CIO only because they were on on good terms, right?
Matthew Chiodi: Mm.
Raghav: so we wanted some sort of independence in that case. And we found quite a few, not not quite a few, but we found a couple of dyads where the CISO was reporting to the CIO and had a very healthy relationship. So they still enjoyed you know, direct board access. They were still enjoying Things like autonomy, escalation rights, budget authority, in a way, and even being involved a bit into s into strategy as well. So I wouldn't tell each and every organization that, hey, move this box on the organizational chart and you've solved your cybersecurity governance issues, right? so I would actually ask them whether things like accountability, Decision authority and even the resources are aligned or not. Because the dis because the dangerous arrangement could be saying to somebody that, hey, you own cyber risk, and then you know, you're not even giving them the levers or the tools that they need to change that risk, which unfortunately is the story of a lot of CISOs.
Matthew Chiodi: And this immediately brings to my mind the very well-known statement that is popularized by Peter Drucker, which is that culture eats strategy for breakfast. Because I think everything you're talking about here is really about organizational culture, right? We go back to the archetypes with those, with those three archetypes. My my sense, and from the many organizations I've worked in over the last 26 years. Is that it always comes down to organizational health? I have seen the reporting structure work, even when the you know, years ago the CISO was buried, not you know, might not have even been under it was the CIO and there might have been something else that they were put under. Maybe they were even put under infrastructure years ago, but it was really healthy because the organizational culture, even outside of just security itself, was a healthy culture. And as you mentioned, I've also seen it in organizations where the CISO was a direct report to the CEO. And that wasn't good because again, the in this case, the organizational culture was not a healthy, conducive culture for that. And they thought, well, hey, if we make this a direct report to the CEO, that looks great to our investors. We're really serious about cyber. It's a direct report to the CEO. And but the end of the day, for me, it always seems to come down. to culture. And it sounds like a lot of your research is a big part of that. So you could likely have, and here I'm gonna, this is gonna be my hypothesis based upon what I've what I'm learning today during our interview and based on what I read in the research was you could have a CISO 3.0 come into an organization has got poor culture. And it doesn't matter where they report, they are going to struggle, they're gonna have these stressors. And maybe not a maybe not a CISO 1.0, but maybe a 2.0 could come into a organization that has you know, maybe just a moderate culture and have much more success than that 3.0, right? Because you can have an individual that's got amazing experience, that is highly self-aware, is really empathetic with their team, but there's only so much that that individual can do to fight. a larger culture, especially in a very large organization. It's somewhat different if we're talking, you know, an early stage startup that's got less than a hundred people. But generally speaking here, I think, you know, we're look we're talking about larger organizations. Does that is that am I off? Does that seem to make sense?
Raghav: that makes a lot of sense. in fact that that just caught me thinking that you know, when we talk about CISO one point two point three point zero, it relies a lot on the environment. So, like
Matthew Chiodi: Mm.
Raghav: you said, you know, a CISO three point in some organization might go to another organization and there they are no longer CISO three point they are probably CISO two point or they might be a CISO three point who are struggling, right, just to reinforce their position and and even vice versa. So Yeah, that can happen. But I mean it's it's a difficult role, right? I mean every CISO that I've talked to, nobody has played it off saying that, you know, it's an easy role and they know exactly what to do. In fact, one of the CISOs actually told me that if you wake up in the middle of the night and you hear crying, it's either a baby or a CISO. The job is that difficult. Yeah. So
Matthew Chiodi: That's a great quote. I love that. I love that.
Raghav: so yeah, that that just stuck with me. It's it's been I think two and a half years since that CISO told me that stuff and it just it still makes me chuckle.
Matthew Chiodi: That's a good one. that's a good one. So let's go to your the kind of the last piece of research here, which is around you know, the genesis of attacks, right? So most security frameworks, MITRE, etc., they usually describe attacks beginning with reconnaissance, initial access, or execution. Some of your your research though argues that the attack may begin earlier, right? During what you called feasibility exploration. Let's, you know, unpack that a little bit for us. Like what exactly is attack conception and what distinguishes ordinary curiosity or an honest honest mistake from emerging malicious intent?
Raghav: So so yeah this project is still a newer work for us, but the basic idea is pretty pretty simple that most of the cybersecurity frameworks become very useful once your recognizable attack behavior begins, right? So you have, like you said, reconnaissance, you have initial access, execution, and so on. what we are asking is, does the attack begin earlier than that? So before somebody executes an attack, they may be exploring feasibility. So they try something, it doesn't work. they learn something from that failure, right? And then they change direction, they test another boundary, and then they discover that you know one route is closed, but another might still be available. So at some point, that exploration starts converging on a viable attack pathway. and That's what we are interested in when we're talking about attack conception. so now, I mean, the really important caveat is that unusual behavior is not automatically malicious behavior, right? And I think you know, security has to be a bit careful there because somebody can make a mistake. you have people in organizations making mistakes all the time, clicking a phishing email every now and then. just being an honest mistake on that part. but yeah, somebody can make a mistake, they can be curious, they can I mean you can have a developer who can do something really weird. you can have somebody on the research team that can test something which is unconventional. So if you treat every anomaly as malicious intent, you're just going to create, you know, a huge surveillance problem and A huge you know false alarm or or a false positive problem as well. So the distinction you know that we're looking at is not necessarily necessarily just one action. I mean the dis the dist the distinction may be in the trajectory itself, right? So if I have to you know use a quick you know analogy, if you imagine That you know somebody just walks up to a building and accidentally tries the wrong door, right? that tells me almost nothing. But if the same person tries the front door, realizes that you know it's locked, and then walks around the building, tries a side entrance, looks at you know some sort of badge reader, and then goes around to the loading dock. so none of those actions individually prove malicious intent. But then the sequence is beginning to tell you something different, right? So that's basically the idea. So instead of instead of just asking, you know, was this particular action malicious? We are trying to understand what the sequence of actions appear to be converging toward, right? So that's what we are trying to get at. That when we're talking about attack conception, it's really about. being able to understand and recognize you know, the patterns that lead to an emerging story before the exploitation phase becomes the first chapter that, you know, sort of is unmistakable of sorts.
Matthew Chiodi: So I think that makes sense. And it sounds like there, you know, my mind goes to some science fiction movies that have occurred over the last 20 years with you know, various different movies, maybe Demolition Man and there was the one that Tom Cruise was in where the precogs where they could tell what somebody was gonna do before they actually did it. I'm sure you're probably not thinking down that route, but that's what comes into my mind. So let's make this practical, right? Like how would a security team perhaps operationalize that across like identity, endpoint, cloud, control plane telemetry without simply creating more alert noise, right? We know, and and then I of course I'm gonna ask you a question about AI here in a minute, but how might someone operationalize this in their in their security team?
Raghav: Right. so I would say that, you know, the first thing is that the worst possible implementation of this idea would just be to create another thousand alerts, right? Because you have security teams who have plenty of alerts, you have you know, a lot of telemetry. I mean identity systems see something, endpoints tools endpoint tools see something, something, and you have cloud systems that Are seeing something else entirely, right? So the control plane sees another piece. And the problem is that we often examine these things separately. So you know, if you have an individual action, that could look something it could look sort of benign. So what gets interesting is that mean it's it's very interesting when you reconstruct what happens over time. So rather than simply asking that, hey, was event X suspicious? You might you should rather ask that, okay, what has this identity been doing across systems for the last hour? What did it try? What failed? What changed after the failure? And you know, where did it go next? So to me, that's moving from just event detection to narrative detection. And actually this This really just, you know, connects directly to another project that I'm working on, which you know, I've sort of coined as security infobesity. so security has had an you know information overload problem for years. Like, you know, you have too many alerts,
Matthew Chiodi: Yes.
Raghav: too many vulnerabilities to look at, too many dashboards and whatnot. Right. So one of the big promises of AI was that, hey, great, we'll fix this. We'll have AI summarize it. AI will help you prioritize it. AI will tell you what matters. And it probably will help with that. But then there's a bit of a paradox that I find really interesting that, you know, AI might reduce the amount of information within a security professional. I mean, that a security professional has to has to consume while increasing the amount of explanation work that they have to perform. So suppose an AI system tells me that. hey, I looked at all of these 30 odd events and I believe this represents an emerging attack pathway. Okay. then my, you know, my mind gets into questions like, okay, why? Which are the events that mattered? Which ones did you infer? How confident are you? And whatnot. So and now, you know, imagine if I'm the CISO and I have to take that judgment to the CIO, to the board, an auditor, or even to a regulator. I cannot necessarily say that, hey, the AI said so. Right. So exactly,
Matthew Chiodi: Yeah, you still have to explain it.
Raghav: yeah. So that's the explanation load. so you may solve one problem, which is the information overload, but then you're inward inadvertently creating another one, which is the explanation overload, right? And that's an interesting connection back to our burnout research here. That you know, one thing that we saw there was that. you know, reducing one stressor doesn't necessarily depend ne necessarily mean that, you know, the burden disappears. So sometimes another stressor substitutes for it. AI could do exactly that. So it could, you know, just remove some of the work of finding the signal, but then it's still creating that new work around verification, interpretation, accountability, and explanation, especially if you know, sometime down the line there's an audit, right? So when I talk about narrative detection, I definitely don't mean just you know throw an LLM on top of the you know SIEM and ask it to tell us scary stories, right? So the real question is that can you have the AI compress the complexity without simply transferring the cognitive burden downstream to the human? Right. And Agents make all of this much more urgent because I mean a human attacker explores at human speed, but you have an agent which can potentially try something very, very quickly, right? And some of that behavior can look very similar to you know, like legitimate automated behavior. So that's why cases which involve agentic systems, including the kind of you know, incidents that you know we are now looking at involving systems such as you open AI and hugging face right so that's why they are so interesting so the questions I mean the questions now become less about trying to infer some psychological notion of intent right and I I don't necessarily need to ask whether an AI wanted to attack something. right. So what I care about is What is the observable behavior? Is the system exploring? Is it learning from beh from failure? Is it adapting? Is it moving towards, you know, some sort of a feasible pathway? So I think we are going to have to get much better at recognizing that pattern or you know, whatever pattern that we are observing. because we are sort of entering an environment where, you know, the the machines can generate behavior faster than humans can interpret it. And then
Matthew Chiodi: Yeah.
Raghav: we may ask other machines to explain that behavior back to us. So that creates a very different security challenge. So I I I actually believe that you know the future problem could I mean it it may not simply just be tackling too many alerts. It may it may be too many machine generated interpretations requiring human judgment. And I think that if we are not careful, we could replace the alert fatigue with explanation fatigue. which I don't I mean, I I don't know if the industry or even the academic research has an answer for how to tackle that.
Matthew Chiodi: I think you're right. It's still, you know, the rate of innovation is continuing to continue to accelerate. There seemed like maybe six to twelve months ago that there might have been a plateau, but I don't think there really was. And so I agree with you, a lot of these things are will take time to solve, and AI does not solve these things, and AI does not remove the need for strategic human thinking and decision making. I think it actually increases. need for it because what I've seen many times is and now I can tell you certain facts but it you don't know if something is a right decision or not. You can't say like you said, well the AI said, right? Try taking that to your board, to your CEO, the AI said, and that's why. Right. So I I I definitely see the the interconnects here between your your research where we were talking about those stressors with Yeah, contingent stressors, complementary, and substitutive stressors. I think AI probably falls into substitutive stressor category for many CISOs and security teams at this point in time. So, so pulling these these studies together, right? Again, they're they're very diverse in terms of what they cover. Give us maybe two or three actions that you would ask a CEO and a CISO to take during the next call it 90 days. What would what would those steps be? And and then also the follow on to that would be what is like one common practice you would tell them to stop.
Raghav: Okay. so I'd probably you know focus on the I mean the first thing that I would focus on is just having the CEO and CISO sit down together and do a very simple accountability versus authority audit, right? So for every major cyber risk the CISO is expected to own, they have to ask what authority does this person actually have to manage it? Do they control the people, do they control the budget? do they even have the power to stop or modify a risky initiative? And how quickly can they escalate something directly? Right. So if the answer is that, you know, hey, you're responsible for this, but somebody else controls all the decisions, then you've found yourself a governance problem. Right. the second thing would be that I would look for recurring organizational friction. Not just technical vulnerability. So you know, asking questions like, okay, where is security always getting in getting involved too late in? which is the resource argument which happens every six months between either the CISO and the CIO or CISO and the CEO? and you know, where does the CISO spend a lot of unreasonable amount of time translating or defending something that should already be institution institutionalized. Right. So if you have all of those patterns, I mean, I think those patterns tell you quite a bit about the security maturity of the organization and, you know, how do they perceive security? And the third one would just be to you know, just take one recent incident. So for example, the hugging hugging phase one. Right, or even a tabletop exercise and reconstruct it as a story rather than just a collection of alerts. So asking questions like, okay, what happened first there, what happened next, what changed, what sort of systems had pieces of the story? And you know, maybe even asking a very uncomforta uncomfortable question at the end that okay, at what point could we reasonably have known that something was beginning to form, right? So I think such questions and having such sort of conversations that starts preparing the CEO and even the organization for you know the kind of pre-attack and agentic behavior that we were just talking about. And if there's one thing that I would tell organizations to stop doing, it would be that you know stop giving CISOs accountability without Them having enough authority to make that change, right? Because we have created such a strange situation in cybersecurity where one executive can sometimes be professionally or publicly and even potentially legally accountable for enterprise cyber risk, while a huge portion of the technology budget, human, I mean, when I say human, I mean personal. And even the business decisions can just create you know that risk. All of that, all of those facets are just con are just being controlled by somebody else or controlled somewhere else. So that in my mind is not s you know, sustainable governance. so yeah, these are some of the things that I would keep in mind. Yeah.
Matthew Chiodi: Rocov, your your research is absolutely thought provoking, which is why I love it. Now, I would imagine someone in your field, you've absolutely need to be on top of, you know, what's happening, other academic research, what's happening in the real world outside of it ac academic research. What is your kind of daily, weekly routine look like? How do you stay sharp?
Raghav: So I mean so I recently graduated from my PhD program. So I have a bit of you know free time between you know now when when I start as an assistant professor in University of Missouri at St. Louis. So right now my my schedule is fairly flexible, but usually you know, I think I I try to stay curious more than anything. I try not to spend all of my time inside the academic. cybersecurity literature. Because you know, one of the pet peeves that I have about you know, academic research, and I'm not trying to be very critical of this, but it's just that it takes a lot of time to publish. So just to just to give you you know, an insight into this, usually a very good research project in academia, it takes you know, from From ideation to it being published, it usually takes at least two to three years. And when you consider the industry which is moving so fast,
Matthew Chiodi: Yeah.
Raghav: right, by the time it gets published, it may not be as substantial as it otherwise would have been. Right? I mean, you could have indust industry people looking at academic literature and saying, Hey, we already know this, what's new about this? But they don't know that, you know, the ideation and everything started maybe. Two years, three years, or even sometimes six years, six years ago. right. So so that's one of my pet peeves. So that's why I try to always look at you know, white papers and you know, the cybersecurity news as well. but yes, obviously as a researcher, I read the research. that's really important to me. But a huge amount of what I learn, like I said, it really comes from talking to people who are actually doing the job. So I've been fortunate that throughout my you know PhD journey, I was fortunate to talk to a lot of security folks, including CISOs. So in all, I've been able to conduct close to 130 CISO interviews. And even though at first it was purely because of research, I learned very quickly that an hour spent interviewing the CISOs was a great learning opportunity for me personally that I couldn't get from anywhere else. Right. So it was really just a privilege to do that. I mean, I had CISOs who would tell me very quickly when something academics think is important isn't particularly important in their day to day life. Right. Yeah.
Matthew Chiodi: In their world. In the real world.
Raghav: And you know, they they also told me about something which consumed half their week that academics barely have a language or a theory for yet. So that was really fascinating to me. So what I try to do is that I try to get information from all of these different sources and I try to triangulate three things. like you know, what academic literature is telling us, what practitioners are experiencing, and what we are actually seeing in actual incidents. And then I read outside cybersecurity quite a bit, like organizational behavior, psychology, leadership, and information systems. Because I've I sort of realized that a lot of supposedly new cybersecurity problems are not entirely new problems. They are you know, sort of old questions about power, trust, accountability, and you know, uncertainty and even human judgment, which are just appearing in a new technological environment. And you know, the last thing is probably just being comfortable saying, I don't know. Because cybersecurity is moving too quickly for anybody to genuinely understand everything that's happening. So for me, saying that I mean, so for me, just staying sharp isn't having to, you know, answer isn't to have an answer for each and everything. so what I what I try to do best at is you know, just just try to do my best at noticing when the answers that we already have stop explaining what we are really seeing. So that's usually where the interesting research begins. so I think that was a long winded way of explaining, you know, where my head is at and how I try to stay I wouldn't use the word sharp. I just stay a bit above the surface. but yeah that that's what I tried to do. Yeah.
Matthew Chiodi: I love that. I love that. Well, I always ask this question of all my guests because, you know, I'm fascinated by the responses. But if I was to just do a quick summary of what you said, it sounds like for you, staying sharp is about regular interaction with those that are outside of academia, right? Who are the subjects of your research, spending time with them? And I think, you know, you know, most of who listens to our podcast is not in academia. And so I think the takeaway for them is if you're not having regular conversations with your colleagues outside of cyber, you just might be a CISO 1.0 or 2.0. So that's my that's my takeaway. So have a lot of those
Raghav: Fair enough. Yeah.
Matthew Chiodi: conversations. It's one of the reasons, quite frankly, people ask me, you know, why are you still why do you still do the podcast? It's been, you know, almost five or six years. This is one of those reasons that I do the podcast. It's because I learn so incredibly much and I try to have on many different voices, mostly yes, cybersecurity, but even outside of it. So Ragav, this has been a tremendous conversation. Thanks for coming on the show.
Raghav: Thank you so much, Matt. I had a lot of lot of fun chatting with you. and yeah, it's it's a privilege. Thank you so much for having me.
Matthew Chiodi: Thank you.
Raghav: Okay.