Pod Insights a16z Podcast a16z Podcast | Barbarians at the Gate -- How to Think About Enterprise Security Today About this episode Enterprises large and small run their applications and infrastructure at a whole new level of agility and speed. But unfortunately, security doesn’t like speed. “The faster you go, the harder it is to understand what is happening and to protect your infrastructure,” says Andrew Rubin, CEO and co-founder of cloud security startup Illumio. So then how do we rethink the architecture of the past to acknowledge the way business happens today? If you want to start tackling the shifting landscape of business and security today, “Go become a student of the economics of war and crime,” suggests Gaurav Banga, CEO and co-founder of Bromium. If going slow is not an option, what can and should we do?
Transcript 0:00 Speaker 2 Welcome to the A16Z Podcast. I'm Michael Copeland, and we are continuing our discussion of security, and we are lucky to have Andrew Rubin, CEO of Illumio and co-founder, and along with Andrew, Gorf Banga, CEO and co-founder of Bromium. Welcome, guys.
0:15 Speaker 0 Welcome. Thank you so much. We are happy to be here.
0:18 Speaker 2 Thank you very much. Gorf, I you I read something that you had had said or written. The barbarians are at the gate. Am I being attacked?
0:28 Speaker 0 Yes, you are. And what barbarians at the gate means that uh never before have we had so much online. We have computerized our every existence, every aspect of our existence, how we invest, how we get paid, how we do healthcare, how do we deliver power, everything. And unfortunately, we have built that on a security platform which is not architecturally sound. And we're getting attacked every day. It's as if war and crime just came online.
1:01 Speaker 2 Andrew, how do you view that and how do your the folks that you talk to kind of internalize the fact that if I have stuff out there that's valuable, people are gonna want to go after it?
1:12 Speaker 1 So I I completely agree with Garov that we're we're effectively digitizing everything, and it's literally everything. It's uh everything from the way that we bank to the way that we hail a taxi or a car to move from point A to point B. So inherently there's a lot more digital and electronic to protect. I think the the the aha moment for security, and it's recent, it's measured probably in months or maybe a year or two, is that this concept of being in a binary state of safe or breached is no longer a viable way to look at the world because with this much out there, it's almost an assumption that you've already been breached or you will be breached and you may not know it right away. And what we're hearing more and more now is this concept of how do I reduce the surface area of attack when I'm breached.
1:57 Speaker 1 That's a very different security conversation than the one that we've had for the last 20 or 25 years.
2:03 Speaker 2 So you're saying it was a matter of months before that kind of mindset shift happened. Why, why finally do you think that occurred?
2:11 Speaker 1 I I I don't think it's any one thing, I think it's a combination of a few things. So the first one is that there's a lot more places to put stuff. I mean, if you think about five years ago, whether or not we really would look at the public cloud as a truly viable alternative to your data center that you built, owned, controlled for the 20 years before that. There was a debate. Now there's no longer that debate. It doesn't mean everything will land in the public cloud, but it means that it's a viable alternative. So we're more distributed and more heterogeneous than we've ever been. I think the other thing that's going on is there's a shift in the way that enterprise is thinking about running their infrastructure and applications, and the shift is all based on agility and speed. And unfortunately, security doesn't really like speed. Those are two things that traditionally have been at war with each other.
2:57 Speaker 1 The faster you go, the harder it is to understand what's happening, and certainly the harder it is to protect it. Unfortunately, that that friction point is no longer tenable. Right. Enterprises are going to go fast and they're going to need to do it with sp with security at the same time. So Garov,
3:13 Speaker 2 for you guys, how do you address that tension between going fast operations and you know security? We've been talking about this, how you know you need to respect security, but you need to get things done. So in that speed, uh in companies in an environment where speed is of the essence, how do you reconcile those things?
3:34 Speaker 0 So to be able to reconcile, you first take a step back, and just to add to what uh Andy said earlier, you know, so the world has changed, the world is changing. Uh you start looking at besides so cloud is one very important development that has happened, another development that has happened is uh mobile, which we all know, right? And then another one that has happened is that you're relying more and more on the internet, which is you know not just cloud as in as in you know you do cloud computing, but the fact that you're very content uh content dependent, you're c you're generating large amounts of content, and you're exchanging and sharing that content, you're trusting each other over the internet. So if you look at all these trends, the first thing you do is you take a step back and you examine how the security architecture must change.
4:24 Speaker 0 And one of the other requirements comes along is that the security architecture must also be responsive to the need to go faster.
4:32 Speaker 2 Right.
4:32 Speaker 0 Now you came up with a set of requirements. You must your new security architecture or your modification to the existing security architecture must be must have these properties. It must deal with cloud, it must deal with mobile, it must deal with consumerization, it must deal with the fact that we are relying more and more on internet content. It must deal with the fact that change is more common. Now then it becomes a computer science, a computer architecture, software design problem. And it it turns out that it is, I mean, we we we live the message of hope. It it does turn out that uh you know it is possible for human innovation to come up with such a design, which is a more sophisticated, a more well-thought design on security, but you can put it just together.
5:15 Speaker 1 And I'll build on that. I just want to kind of add one thing that you know we when we launched Illumio into the market last October, so about six months ago, obviously the amount of feedback that we started to get because we were talking to more people and certainly talking more openly went up very dramatically. And one thing that's interesting is consistently across the board, customers are saying to us that they're finding that there isn't a natural or easy iterative path from the architecture of the past. Well, Gorov said about having to rethink the problem from first principles. We're actually hearing customers say that. So despite the fact that you started off mentioning that everybody's dressed in black, and so it must be security. What's interesting is that it actually doesn't seem that eerie for one very simple reason. Because for the first time in decades, the customers are actually in a place where they're willing to truly rethink this from the very beginning.
6:03 Speaker 1 They understand that there's a new set of problems and a new set of challenges that security has to face that aren't built on the problems of the past, and therefore they're willing to look at a completely new way of solving it. That's a massive change in the enterprise of the customer mindset.
6:18 Speaker 2 That that goes chicken jowl, I guess, with with this shift to the cloud, right? I mean, they're they're willing to look at that in terms of running a business, and so they're also willing to look at ways to change their security approach.
6:29 Speaker 1 Well, and I think if you look at it, some of the organizations that you would think are least likely to take advantage of things like public cloud or allow open access through mobility, if they're willing to do that, then it's not a leap or very far step to imagine them being willing to look at security through a completely new lens for the first time in a long time. The challenge is that the industry has to respond by bringing things to market that actually start from a blank page and allow the customer to look at it not only as a new set of problems, but also from a completely different way of trying to solve them. And so we have an obligation sitting on our side of the table, Garov and I and and others, to actually bring things to the customer that fundamentally start from a different place than just simply iterating on the architecture or the model of the past.
7:14 Speaker 1 So
7:15 Speaker 2 you guys have a different philosophies in your companies about how to approach all this change. Um barbarians are at the gate, they're trying to get in all sorts of different ways in new ways all the time. How do you then anticipate kind of the new? You know, it's one thing to change my architecture and and sort of head off in a new direction, but if I don't know where the next breach or attack or you know, bad thing could come from, how do I how do you approach that? So
7:44 Speaker 0 I mean, this is this is if you take a step back, and and it's hard to take step back because it's just life is so busy. Um we take a fresh step back. The problem that we are trying to deal with, which is what you were terrified about, unseen and unknown, what you don't know, what you don't see, what you don't see coming. The instructive way to think about it is actually just go back to the drawing board again and say, well, what has happened? Two things have happened. The way we do IT is changing with cloud and mobile and all of that. That's one aspect. The other aspect is we have so much online. Just forget so imagine we had none of that IT change. We still have so much online that it has become very rewarding for the bad guys or the adversary to come back and after you in the online space, right?
8:33 Speaker 0 Now, this is none of this is new. So shifts in IT have happened before, shifts in our way of life have happened before. And war and crime is as old, older than mankind, as old as mankind, whatever, right? Do you want to look at it? So the way you want to think about this is go become a student in war and crime, what how war and crime works, what are the economics of war and crime, and then go become a student of some of the computer science behind that gives you the approach you need to take. That gives you the approach. So for example, why why would people come in and say, I'm gonna spend 10,000 dollars on just buying this software exploit so that I can hack this Fortune 500? Why? Because that 10,000 is a small fraction of the reward that you would earn from that.
9:21 Speaker 0 And it is much, much cheaper than trying to attack the bank in the physical world. Right. That's the reason why they do it. Right. What what makes the job easier? What makes the job easier is the sheer complexity of IT, but also the fact that things are shifting in IT security is behind the shift, like the fact that you have cloud, the fact that you have so now the best approach is first to recognize that this is happening and then to come back and design what your response to this is gonna be.
9:53 Speaker 2 Andrew, how do you guys approach that? I know you you you talk about reducing
9:58 Speaker 1 surface area. We do. So we Talk a lot about reducing the surface area of attack because there's a premise that security functioned in a very binary world for a long time. Security's job was to keep you safe. And safe inherently meant that nothing was wrong. And of course, when safe fails, then it seems like everything is wrong. The way we would say it is, it felt like you were either perfectly safe or catastrophically breached. And what we're finding is that what customers are now working off of is just a fundamentally different assumption, which is I'm probably breached. If I'm not already, I will be. And it's equally interesting and maybe even more so to ask the question when that happens, what is the surface area of attack? How much damage will something inflict? What is the blast radius inside of my data center or cloud when something goes wrong?
10:43 Speaker 1 So from an Illumio perspective, we really look at it in terms of mirroring the compute environment, the infrastructure and application environment, so that security doesn't feel like a bolt-on, doesn't feel like something that gets tagged on after the fact. But security is from the very beginning built into the infrastructure and the applications and follows the motion as things drift and change over time. And part of our story is to distribute the policy and the enforcement out to all of the individual workloads so that the surface area of attack is no longer the perimeter or all the things behind it, but the individual workload itself and how it's talking to and communicating with other things inside of the environment. So you
11:22 Speaker 2 get access to just this small slice, you know, if even that.
11:26 Speaker 1 That's exactly right. And actually what's interesting is even the perimeter in its most traditional sense, when we used to wrap a brick wall around an entire data center, that really was effectively the same theory in that I was putting a brick wall around a group of assets, a bunch of servers sitting inside, and therefore they were protected. What we're doing is we're simply taking that and shrinking the surface area of attack down dramatically, dramatically to the point where it could be a single server, a single VM, and now with an announcement that we made last week, even a single process running on one of those compute instances. But all of it comes back to the same thing. How do we have the ability to distribute security dynamically, make sure that it's always provisioned correctly in a dynamic world, and how do we reduce the surface area of attack?
12:11 Speaker 1 Garf,
12:12 Speaker 2 do you guys again there's this idea that, well, it's security, it's going to slow me down, it's going to be a pain in my arse. You know, how do you make sure people use it? And how do you advise, you know, your customers and and um folks in this world to make it easier?
12:28 Speaker 0 So actually, you know, the the thing that makes it easier is when you design with these assumptions built in. When you design something, the mobile is not excluded, the internet is not excluded, the cloud is not excluded. And some of the tools that you use are, you know, this whole idea that Andy talked about earlier, which is micro-segmentation, micro-virtualization, which is what Chromium does, whether you do it in the network and the data center, or like what does in the endpoint, um, is it it's it's it gives you that exact tool. So why what do people care about? People care about doing whatever they want to do. That's really what it is. Right? They want to click on everything. Right. They have CDL clickers, if you will.
13:10 Speaker 2 Right.
13:10 Speaker 0 Right? So if you want to click on anything, if you want to run whatever you want to run, and you cannot be told that you may not do that, then the question really becomes is how can we create the environment and the infrastructure so that you can do that safely?
13:22 Speaker 1 And
13:22 Speaker 0 the approach that Bromium takes, the approach that we take and uh a whole bunch of other, and it's not very dissimilar, it's actually a dual of what Andy just talked about, uh, but just from an end user perspective, but when you're running a piece of code and you don't know about the origins of that piece of code, that it could be a website or whatever, one thing would be to give it, give the website access to your entire computing environment. Another way would be to create a virtual machine container in which the website is allowed to run, and this thing may not escape, whatever, whatever the side effects of this website are not allowed to escape the virtual machine container. Now, this is very powerful because you never say no. Right? Virtualization allows you to build boxes, tiny boxes around untrusted pieces of competition.
14:07 Speaker 0 That means you never say no, you always allow any kind of competition. You just build boxes that control what leaks out of that container. What is the scope of that competition? So this becomes very empowering because in our system or in the system that are built in this way, you can literally do whatever you want to do. It is just that when you are it's like using uh burner cell phones, right? You use it by cell phone, you throw it away. It's like using disposable gloves. So if you have a thick enough disposable glove, you can touch anything. Why? Because you don't really care. It's gonna get dirty, you're gonna throw it away. Right. Right? But it gives you this power of being able to touch the dirtiest of things and the sickest of patients and and so on and so forth. So this is a very different paradigm shift where you're designing the infrastructure from the ground up in such a way that saying no to the end user is not an option.
14:57 Speaker 0 You are going to be secure in spite of the user being able to do one thing to do anything and click on anything.
15:03 Speaker 2 And go as fast as they want. And go as fast
15:05 Speaker 0 as
15:06 Speaker 2 they want.
15:06 Speaker 0 I see.
15:07 Speaker 2 Um let's talk about courage versus uh foolishness and take a step back. You know, I'm gonna have the courage to move to the cloud as a as a company as a potential customer. You know, what's courageous and what's foolhardy? You know, courageous is I don't know what, but foolhardy is staying on Windows XP, for example. Um how do you guys view that?
15:27 Speaker 0 So it really is, you know, there is a there's one of our friends, uh, the CISO of Aetna, Jim Routh, he said just very famously, you know, this is the top, this is the 10%. What that really means, it's a CISO or the SISO, depending on how you pronounce it. The chief information security officer, that takes risk to reduce risk. Here's the reality. The world is changing. If you say status quo, you might think that your risk to your business is not is not increasing. The reality is that it's increasing really, really, really fast. Right faster than you can control. So in order to deal with the risk, in order to deal with the changing conditions, you have to take a risk. And unfortunately, none of the existing big vendors is going to give you what you want.
16:13 Speaker 0 You have to take think about a new approach. So foolhardy is going through the world thinking, nothing has changed, it's business as usual, I can keep saying no to the end users until the such day that my company is going to get breached or I'm going to get fired or I'm going to have to fire somebody.
16:32 Speaker 2 Right.
16:32 Speaker 0 Or smart is realizing that things are changing, go through this process of selecting and deciding what is good, what could be good, taking them to the basis and moving.
16:48 Speaker 1 I I'll add that I think uh I think courageous in this case is actually responding to the needs of the organization and being able and willing to look at any tool, any form of infrastructure, any operating model that allows the enterprise to do what it needs to do. It takes courage to actually say we're going to implement completely different things than we have in the past, but we're doing it because the business requires it. I think foolhardy is assuming that the only thing that you have to secure all these new things is what you've had in the past. And what we're finding, like I said earlier, is that customers for the first time are actually very open to looking at completely new and different things because they realize that they're solving for a new and different set of problems.
17:28 Speaker 2 Let's talk about mobile a little bit. Garf, you brought it up and Andrew, you you've referenced it. Um mobile malware, everybody's got a smartphone, not all of us, you know, snap them in half and throw them away after we're we're done with them. Um what's new in the mobile world and what are you guys seeing and how people how are people responding?
17:46 Speaker 0 So our view of mobile is uh slightly nuanced than that. That is that is there's a laptop and a tablet, which is a real mobile vector. And it's it's it's a vector of attack, primarily because it leaves the four walls of the enterprise and all of the traditional defenses which rely on firewall and those are just out of line.
18:05 Speaker 2 Right.
18:05 Speaker 0 Right. So that's the reason why these things are far easier to get to, far easier to attack, and and and you know, just that's just the economics of it. Uh the world of mobile smartphones introduces or and also tablets introduces a different kind of problem, which is the problem of information management, right? These things are bring your own devices, consumerized devices, they're very primitive controls in terms of information management. Uh while there's not much malware exists, some malware does exist, but not much malware exists for attacking mobile devices themselves, like in the Android and the iOS case. But the more important thing is that the CIO has very little visibility and very few controls into what's actually going to happen. Now there are companies that are doing the right things and providing you with the right levels of control, and and CIOs, some of them are trying to adopt those controls and and being successful, of course.
18:59 Speaker 0 Uh a lot more work needs to be done. And
19:01 Speaker 1 what we're finding is that it's becoming, in a sense, just another access point. And the reason I say just another is not to diminish how important it is to understand that I think every CIO's dream would be to really truly be able to have a perfect picture of everything that can access every application and every piece of data in their environment, no matter where it is or who provided it, but it's it's in a sense it's a fool's mission. Um number one, because it's hard to drill that kind of control over an organization nowadays, and number two, because it's somewhat antithetical to the way that the business is trying to operate to enable speed and agility. What we're finding is that what customers are doing is they're actually identifying what they consider to be their highest value targets. They're identifying the applications and specifically the data that is the most important asset that they have, they're figuring out where those things are, and they're realizing that they have to protect those things at really all cost.
19:51 Speaker 2 Wherever it is, however it gets accessed, I assume.
19:54 Speaker 1 That's
19:54 Speaker 2 exactly right. Um what about taking the offensive? We talk a lot about gates and walls and perimeters, and we know that those are being Breached, but what about like rushing out and going after folks or making sure that the attacks don't even happen in the first place?
20:06 Speaker 1 I I I there would be there would be an interesting assumption if you could actually work off of the premise that the attack doesn't happen in the first place. I can only tell you anecdotally that in the customer conversations I'm having, they're certainly coming at it from the opposite angle, which is the attacks are not only persistent, they're not only growing in number and frequency, but in a lot of cases they're growing in severity. And so I think the question is what is the definition of proactive if that's the premise of the question.
20:35 Speaker 2 Right.
20:36 Speaker 1 And what we're finding is that the the definition of proactive is to actually understand exactly what it is that you have, where it's running, how these things are talking to each other, and then put a set of controls in place that actually allow you to ensure that the right things are happening, and thereby the wrong things, if and when they do happen, are immediately flagged as out of profile and are either stopped or certainly responded to very quickly. Right.
21:02 Speaker 2 So it still doesn't sound like you know, in in that sort of offense versus defense kind of view of the world, you need to get you know your processes in place, know what you have, know what you're securing, know what's important, and then once that's done, and maybe that's never done, um, think about going after somebody or or or some next phase of security, I guess.
21:23 Speaker 1 Well, and I'll add one other thing, which is that what we're finding more and more often now is that visibility leads to knowledge and knowledge actually allows you to secure whatever it is that you're trying to protect. Um there's been a lot of security thrown at a lot of organizations without really truly understanding what it is that it's protecting. Because as Garov mentioned a few moments ago, you know, the world's gotten not only very scaled, but it's also become very dynamic and very complex. And so the what seems to be simple task of simply understanding where are all of my assets in the Illumia world, it would be where are all my compute instances how are they talking to each other. That's not a static problem any longer. It's not a snapshot where you look at it at Monday morning at eight o'clock, and that picture remains resident for the next two months or six months or two years. That picture actually looks different 15 minutes later.
22:10 Speaker 1 And so just understanding, being able to see and understand what's happening. If you have that, you're probably going to do a much better job protecting yourself. But that's a very big challenge before you ever get to the protection piece of the story.
22:23 Speaker 0 Yeah, I agree. I mean, attribution and then going after the bad guys, I think we have ways to go. And and it just what we can do uh in terms of uh, you know, it's just very easy to misattribute something to somebody today. Um it's I I think we have a lot of technical work to do in that. And then also uh we have we we have very primitive controls across countries. We don't have the interpol equivalent, if you will. Right. Um we don't have the nation-state to nation state agreement that this is not done and this is V gonna, you know, extradite those people and bring them to a foreign jail if they do X, Y, and Z. The legal systems around cybersecurity are much more improved now than 10 years ago, but they're still very primitive compared to, you know, murder and you know, physical extortion and physical theft and glandular city and all of those guys things.
23:12 Speaker 0 Um I think we have a ways to go uh before that will happen. And maybe nation states can do this to each other, but um I I doubt whether commercial enterprises should go over there. Uh they have the means to go over there successfully.
23:26 Speaker 2 Right. So so based on what you guys have told me and and discussed, uh it doesn't sound like there's a we should all light our hair on fire and go running in the streets, no need to get hysterical. But if I'm uh chief uh security officer, if I'm a CEO, if I'm uh, you know, on a board, what is, you know, if we can't win against these attacks, what can we hope for and what what does winning sort of look like? You know, and I'm doing air quotes around winning if if it's not beating them. I
23:55 Speaker 1 think winning is enabling the organization to do what it needs to do to conduct business to remain competitive to grow, and there's a whole series of things that we've done with the infrastructure and the applications and really the entire IT model to allow that to happen better than it ever has. And then security has to realize that its job is to protect that motion no matter what it looks like. And so the reality is the answer for what is security look like today is probably going to be different even a year from now, and certainly five years from now. So it's not a fixed answer. It's not that there is the right model and the right model is the only model. The model is that security has to evolve as quickly and as dynamically as the infrastructure and applications that it's protecting. And so long as those things keep changing, security had better find a way to keep up and mirror it.
24:40 Speaker 1 I I think
24:40 Speaker 0 change is the answer. You have to what is what is what is very obvious is the existing way of doing anything doesn't work. And if you're not embracing change, the right kind of change, empowering someone who is, you know, got a lot of budget and money behind them, a good clear charter to see what you want to do first and to see what we're gonna do second, then you're not doing your job as CCO of uh a global company.
25:05 Speaker 2 Well, uh change is gonna happen and uh risk needs to be taken in a smart way, it sounds like so uh we'll keep an eye on this and we'll keep talking to you guys. Garov, Andrew, thank you so much. Thanks so much for having
25:16 Speaker 0 us.
Previous Company Building in Crypto: A Conversation between Brian Armstrong and Chris Dixon Next What's Next for Education Startups Page updated August 25, 2026 · transcript and summary are machine-generated