Join HP's Chief Information Security Officer (CISO) to learn about how some of the very largest global enterprises like HP are exploring all of their options for doing business safely and continuously.
At HP for approximately eight months, Wahlin previously put the security in place after the infamous PlayStation breach while he was the chief security officer (CSO) at Sony Network Entertainment. Prior to that, he was the CSO at McAfee, after a stint as CSO at Los Alamos Laboratory. Years ago, Wahlin got his start doing counterintelligence for the US Army during the Cold War.
Below are some excerpts.
Wahlin: Big data is quite an interesting development for us in the field of security. If we look back on how we used to do security, trying to determine where our enemies were coming from, what their capacities were, what their targets were, and how we're gathering intelligence to be able to determine how best to protect the company, our resources were quite limited.
We've found that through the use of big data, we're now able to start gathering reams of information that were never available to us in the past. We tend to look at this almost in a modern-warfare type of perspective.
If you're a battlefield commander, and you're looking at how to deploy defenses, how would you deploy those offenses, and what would be the targets that your enemies are looking for? You typically then look at gathering intelligence. This intelligence comes through multiple sources, whether it's electronic or human signals, and you begin to process the intelligence that's gathered, looking for insights into your enemy.
This could be the enemy's capabilities, motivation, resourcing, or targets. Then, by that analysis of that intelligence, you can go through a process of moving your defenses, understanding where the targets may be, and adjusting your troops on the ground.
Big data has now given us the ability to collect more intelligence from more sources at a much more rapid pace. As we go through this, we're looking at understanding these types of questions that we would ask as if we were looking at direct adversaries.
We're looking at what these capabilities are, where people are attacking from, why they're attacking us, and what targets they're looking for within our company. We can gather that data much more rapidly through the use of big data and apply these types of analytics.
We begin to ask different questions of the data and, based on the type of questions we're asking, we can come up with some rather interesting information that we never could get in the past. This then takes us to a position where that advanced analytics allows us to almost predict where an enemy might hit.
That's in the future, I believe. Security is going from the use of prevention, where I'm tackling a known bad thing, to the point where I can use big data to analyze what's happening in real time and then predict where I may be attacked, by whom, and at what targets. That gives me the ability to move the defenses around in such a way that I can protect the high-value items, based on the intelligence that I see coming in through the analytics that we get out of big data.
Muller: Brett, you talk a lot about the idea of getting in front of the problem. Can you talk a little bit about your point of view on how security, from your perspective as a practitioner, has evolved over the last 10-15 years?
Wahlin: Certainly. That's a great question. Years ago, we used to be about trying to prevent the known bad from happening. The questions we would ask would always be around, can it happen to us, and if it does, can we respond to it? What we have to look at now is the fact that the question should change. It should be not, "Can it happen to us," but "When is it going to happen to us?" And not, "Can we respond to it," but "How can we survive it?"
If we look at that type of a mind-shift change, that takes us back to the old ways of doing security, where you try to prevent, detect, and respond. Basically, you prevented the known bad things from happening.
This went back to the days of — pick your favorite attack from years ago. One that I remember is very telling. It was Code Red, and we weren't prepared for it. It hit us. We knew what the signature looked like and we were able to stop it, once we identified what it was. That whole preventive mechanism, back in the day, was pretty much what people did for security.
Fast forward several years, and you get into that new era of security threats highlighted by attacks like Aurora, when it came out. Suddenly, we had the acronyms that flew all over, such as APT — advanced persistent threats — and advanced malware. Now, we have attacks that you can't prevent, because you don't know them. You can't see them. They're zero-days. They're undiscovered malware that's in your system already.
Detect and respond
That changed the way we moved our security. We went from prevent to a big focus on not just preventing, because that becomes a hygiene function. Now, we move in to detect-and-respond view, where we're looking for anomalies. We're looking for the unknown. We're beefing up the ability to quickly respond to those when we find them.
The evolution, as we move forward, is to add a fourth dimension to this. We prevent, detect, respond, and predict. We use elements like big data to understand not only how to get situational awareness, where we connect the dots within our environment, but taking it one step further and being able to predict where that next stop might land. As we evolve in this particular area, getting to that point where we can understand and predict will become a key capability that security departments must have in future.
A reminder to our audience, don't forget to follow the HP Protect 2013 conference activities next week, September 16-19.
As I hear you talking about getting more data, being proactive, and knowing yourself as an organization, Brett, it sounds quite similar to what we have been hearing for many years from the management side, to know yourself to be able better maintain performance standards and therefore be able to quickly remediate when something went wrong.
Are we seeing a confluence between good IT management practices and good security practices, and should we still differentiate between the two?
Wahlin: As we move into the good management of IT, the good management of knowing yourself, there's a hygiene element that appears within the correlation end of the security industry. One of the elements that we look at, of course, is how to add all this additional complexity and additional capability into security and yet still continue to drive value to the business and drive costs out. So we look for areas of efficiencies and again we will draw many similarities.
As you understand the managing of your environments and knowing yourself, we'll begin to apply known standards that we'll really use in the governance perspective. This is where you will take your hygiene, instead of looking at a very elaborate risk equations. You'll have your typical "risk equals threat times vulnerability times impact," and what are my probabilities.
It gets very confusing. So we're trying to cut cost out of those, saying that there are known standards out there. Let's just use them. You can use the ISO 27001, NIST 800-53, or even something like a PCI DSS. Pick your standard, and that then becomes the baseline of control that you want to do. This is knowing yourself.
With these controls, you apply them based on risk to the company. Not all controls are applied equally, nor should they be. As you apply the control based on risk, there is evaluation assessment. Now, I have a known baseline that I can measure myself against.
As you began to build that known baseline, did you understand how well you're doing from a hygiene perspective? These are all the things that you should be doing that give you a chance to understand what your problem areas are.
As you begin to understand those metrics, you can understand where you might have early-warning indicators that would tell you that that you might need to pay attention to certain types of threats, risks, or areas within the company.
There are a lot of similarities as you would look at the IT infrastructures, server maintenance, and understanding of those metrics for early warnings or early indicators of problems. We're trying to do the same security, where we make it very repeatable. We can make it standards-based and we can then extend that across the company, of course always being based on risk.
Muller: There is one more element to that, Dana, such as the evolution of IT management through, say, a framework like ITIL, where you very deliberately break down the barriers between silos across IT.
Similarly, I increasingly find with security that collaboration across organizations — the whole notion of general threat intelligence — forms one of the greatest sources of potential intelligence about an imminent threat. That can come from the operational data, or a lot of operational logs, and then sharing that situational awareness between the operations team is powerful.
At least this works in the experience that I have seen with many of our clients as they improve security outcomes through a heightened sense of what's actually going on, across the infrastructure with customers or users.
One of the greatest challenges we have in moving through Brett's evolution that he described is that many executives still have the point of view that I have a little green light on my desktop, and that tells me I don't have any viruses today. I can assume that my organization is safe. That is about as sophisticated a view of security as some executives have.
Then, of course, you have an increasing level of awareness that that is a false sense of security, particularly in the financial services industry, and increasingly in many governments, certainly national government. Just because you haven't heard about a breach today, that doesn't mean that one isn't actually either being attempted or is, in fact, being successful.
One of the great challenges we have is just raising that executive awareness that a constant level of vigilance is critical. The other place where we're slowly making progress is that it's not necessarily a bad thing to share negative experiences.
Wahlin: Absolutely. We look at the inevitability of the fact that networks are penetrated, and they're penetrated on a daily basis. There's a difference between having unwanted individuals within your network and having the data actually exfiltrated and having a reportable breach.
As we understand what that looks like and how the adversaries are actually getting into our environment, that type of intelligence sharing typically will happen amongst peers. But the need for the ability to actually share and do so without repercussions is an interesting concept. Most companies won't do it, because they still have that preconceived notion that having somebody in your environment is binary — either my green light is on, and it's not happening, or I've got the red light on, and I've got a problem.
In fact, there are multiple phases of gray that are happening in there, and the ability to share the activities, while they may not be detrimental, are indicators that you have an issue going on and you need to be paying attention to it, which is key when we actually start pointing intelligence.
I've seen these logs. I've seen this type of activity. Is that really an issue I need to pay attention to or is that just an automated probe that's testing our defenses? If we look at our environment, the size of HP and how many systems we have across the globe, you can imagine that we see that type of activity on a second-by-second basis.
We have to understand which ones of these we need to pay attention to and have the ability to not only correlate amongst ourselves at the company, but correlate across an industry.
HP may be attacked. Other high-tech companies may also be attacked. We'll get supply-chain attacks. We look at various types of politically motivated attacks. Why are they hitting us? So again, it's back to the situational awareness. Knowing the adversary and knowing their motivations, that data can be shared. Right now, it's usually in an ad-hoc way, peer-to-peer, but definitely there's room for some formalized information sharing.
Muller: Especially when you consider the level of information sharing that goes on in the cybercrime world. They run the equivalent of a Facebook almost. There is a huge amount of information sharing that goes on in that community. It's quite well structured. It's quite well organized. It hasn't necessarily always been that well organized on the defense side of the equation. I think what you're saying is that there's opportunity for improvement.
Wahlin: Yes, and as we look at that opportunity, the counterintelligence person in me always has to stand up and say, "Let's make sure that we're sharing it and we understand our operational security, so that we're sharing that in a way that we're not giving away our secrets to our adversaries." So while there is an opportunity, we also have to be careful with how we share it.
Muller: You, of course, wind up in the situation where you could be amplifying bad information as well. If you were paranoid enough, you could assume that the adversary is actually deliberately planting some sort of distraction at one corner of the organization in order to get to everybody focused on that, while they quietly sneak in through the backdoor.
Brett, returning to this notion of actionable intelligence and the role of big data as an important tool, where do you go for the data? Is it strictly the systems, the systems log information? Is there an operational side to that that you tap more than the equipment, more than the behaviors? What are the sources of data that you want to analyze in order to be better at security?
Wahlin: The sources that we use are evolving. We have our traditional sources, and within HP, there is an internal project that is now going into alpha. It's called Project HAVEn and that's really a combination of ArcSight, Vertica, and Autonomy, integrating with Hadoop. As we build that out and figure out what our capabilities are to put all this data into a large collection and being able to ask the questions and get actionable results out of this, we begin to then analyze our sources.
Sources are obvious as we look at historical operation and security perspective. We have all the log files that are in the perimeter. We have application logs, network infrastructure logs, such as DNS, Active Directory, and other types of LDAP logs.
Then you begin to say, what else can we throw in here? That's pretty much covered in a traditional ArcSight type of an implementation. But what happens if I start throwing things such as badge access or in-and-out card swipes? How about phone logs? Most companies are running IP phone. They will have logs. So what if I throw that in the equation?
What if I go outside to social media and begin to throw things such as Twitter or Facebook feeds into this equation? What if I start pulling in public searches for government-type databases, law enforcement databases, and start adding these? What results might I get based on all that data commingling?
We're not quite sure at this point. We've added many of these sources as we start to look and ask questions and see from which areas we're able to pull the interesting correlations amongst different types of data to give us that situational awareness.
There's still much to be done here, much to be discovered, as we understand the types of questions that we should be asking. As we look at this data and the sources, we also look at how to create that actionable intelligence.
The type of analysts that we typically use in a security operations center are very used to ArcSight. I ingest the log and I see correlations. They're time-line driven. Now, we begin to ask questions of multiple types of data sources that are very disparate in their information, and that takes a different type of analyst.
Not only do we have different types of sources, but we have to have different types of skill sets to ask the right questions of those sources. This will continue to evolve. We may or may not find value as we add sources. We don't want to add a source just for the heck of it, but we also want to understand that we can get very creative with the data as it comes together.
Muller: There are actually two things that I think are important to follow up on here. The first is that, as it's true of every type of analytics conversation I am having today, everyone talks about the term "data scientist." I prefer the term "data artist," because there's a certain artistry to working out what information feeds I want to bring in.
The other element is that, once we've got that information, one of the challenges is that we don't want to add to the overhead or the burden of processing that information. So it's being able to increasing apply intelligence to, as Brett talked about, mechanistic patterns that you can determine with traditional security information. Event management solutions are rather mechanistic. In other words, you apply a set of logical rules to them.
Increasingly, when you're looking at behavioral activities, rules may not be quite as robust as looking at techniques such as information clustering, where you look for hotspots of what seem like unrelated activities at first, but turn out later to be related.
There's a whole bunch of science in the area of crime investigation that we've applied to cybercrime, using some of the techniques, Autonomy for example, to uncover fraud in the financial services market. That automation behind those techniques increasingly is being applied to the big-data problem that security is starting to deal with.
You were describing this opportunity to bring so much different information together, but you also might have unintended consequences. Have you plumbed that at all?
Wahlin: Yes. As we further evaluate these data sources and the ability to understand, I believe that the insight into using the big data, not only for security, but as more of a business intelligence (BI) type of perspective has been well-documented. Our focus has really been on trying to determine the patterns and characteristics of usage.
While we look at it from a purely security mindset, where we try to develop patterns, it takes on a counter-intelligence way of understating how people go, where people go, and what do they do. As people try to be unique, they tend to fall into patterns that are individual and specific to themselves. Those patterns may be over weeks or months, but they're there.
Right now, a lot of times, we'll be asked as a security organization to provide badge swipes as people go in and out of buildings. Can we take that even further and begin to understand where the efficiency would come in based on behaviors and characteristics with workforces. Can we divide that into different business units or geography to try to determine the best use of limited resources across companies? This data could be used in those areas.
The unintended consequence that you brought up, as we look at this and begin to come up with patterns of individuals, is that it begins to reveal a lot about how people interact with systems — what systems they go to, how often they do things — and that can be used in a negative way. So there are privacy implications that come right to the forefront as we begin to identify folks.
That that will be an interesting discussion going forward, as the data comes out, patterns start to unfold, patterns become uniquely identifiable to cities, buildings, and individuals. What do we do with those unintended consequences?
It's almost going to be sort of a two-step, where we can make a couple of steps forward in progress and technology, then we are going to have to deal with these issues, and it might take us a step back. It's definitely evolving in this area, and these unintended consequences could be very detrimental if not addressed early.
We don't want to completely shut down these types of activities based on privacy concerns or some other type of legalities, when we could actually potentially solve for those problems in a systematic perspective, as we move forward with the investigation of the usage of those technologies.
Muller: The question we always need to bear in mind here is, as Brett talks about it, what are the potential unintended consequences? How can we get in front of those potential misuses early? How can we be vigilant of those misuses and put in place good governance ahead of time?
There are three approaches. One is to bury your head in the send and pretend it will never happen. Second is to avoid adopting a technology at all for fear of those unintended consequences. The third is to be aware of them and be constantly looking for breaches of policy, breaches of good governance, and being able to then correct for those if and when they do occur.
What is HP is doing that will set the stage and perhaps help others to learn how to get started in terms of better security and better leveraging of big data as a tool for better security?
Wahlin: As HP progresses into the predicted security front, we're one of, I believe, two companies that are actually trying to understand how to best use HAVEn as we begin the analytics to determine the appropriate usage of the data that is at our fingertips. That takes a predictive capability that HP will be building.
We've created something called the Cyber Intelligence Center. The whole intent of that is to develop the methodologies around how the big data is used, the plumbing, and then the sources for which we actually create the big data and how we move logs into big data. That's very different than what we're doing today, traditional ArcSight loggers and ESMs. There are a lot of mechanics that we have to build for that.
Then, as we move out of that, we begin to look at the actual actionable intelligence creation to use the analytics. What questions should we ask? Then, when we get the answer, is it something we need to do something about? The lagging piece of this would be the actual creation of agile security. In some places, we even call it mobile security, and it's different than mobility. It's security that can actually move.
If you look at the war-type of analogies, back in the day, you had these columns of men with rifles, and they weren't that mobile. Then, as you got into mechanized infantry and other types of technologies came online, airplanes and such, it became much more mobile. What's the equivalent to that in the cyber security world, and how do we create that.
Right now, it's quite difficult to move a firewall around. You don't just unplug or re-VLAN a network. It's very difficult. You bring down applications. So what is the impact of understanding what's coming at you, maybe tomorrow, maybe next week? Can we actually make a infrastructure such that it can be reconfigured to not only to defend against that attack, but perhaps even introduce some adversarial confusion.
I've done my reconnaissance. It looks like this. I come at it tomorrow, and it looks completely different. That is the kill chain that will set back the adversary quite a bit, because most of the time, during a kill chain, it's actually trying to figure out where am I, what I have, where the are assets located, and doing reconnaissance through the network.
So there are a lot of interesting things that we can do as we come to this next step in the evolution of security. At HP, we're trying to develop that at scale. Being the large company that we are, we get the opportunity to see an enormous amount of data that we wouldn't see if we are another company.
Paul, it almost sounds as if security is an accelerant to becoming a better organization, a more data-driven organization which will pay dividends in many ways.
Muller: I completely agree with you. Information security and the arms race, quite literally the analogy, is a forcing function for many organizations. It would be hard to say this without a sense of chagrin, but the great part about this is that there are actually technologies that are being developed as a result of this. Take ArcSight Logo as an example, as a result of this arms race.
Those technologies can now be applied to business problems, gathering real-time operational technology data, such as seismic events, Twitter feeds, and so forth, and being able to incorporate those back in for business and public-good purposes. Just as the space race threw up a whole bunch of technologies like Teflon or silicon adhesives that we use today, the the security arms race is generating some great byproducts that are being used by enterprises to create value, and that's a positive thing.
Wahlin: The analogy of the space race is perfect, as you look at trying to do the security maturation within an environment. You begin to see that a lot of the things that we're doing, whether it's understanding the environment, being able to create the operational metrics around an environment, or push into the fact that we've got to get in front of the adversaries to create the environment that is extremely agile is going to throw off a lot of technology innovations.
It's going to throw off some challenges to the IT industry and how things are put together. That's going to force typically sloppy operations — such as I am just going to throw this up together, I am not going to complete an acquisition, I don't document, I don't understand my environmental — to clean it up as we go through those processes.
The confusion and the complexity within an environment is directly opposed to creating a sense of security. As we create the more secure environment, environments that are capable of detecting anomalies within them, you have to put the hygienic pieces in place. You have to create the technologies that will allow you to leapfrog the adversaries. That's definitely going to be both a driver for business efficiencies, as well as technology, and innovation as it comes down.