The problem with signature based security tools is you are vulnerable until the signature is released and distributed. Palo Alto Networks takes a different approach with Traps, so Network World Editor in Chief John Dix tracked down Palo Alto VP of Product Marketing Scott Gainey for an inside look at how Traps works.
Palo Alto VP of Product Marketing Scott GaineyPalo Alto Networks VP of Product Marketing Scott Gainey
You recently unveiled a new endpoint protection product called Traps. Tell us what that’s about.
If I’m outside of my corporate network operating on an unsecured Wi-Fi network my system is at risk. A simple drive-by-download of embedded malicious content in, say, an iFrame could easily bypass existing anti-virus software, leaving nothing that could protect me from being infected. This is one of many examples that leave endpoints vulnerable. So a complete security architecture has to be able to protect its users regardless of where they may be working, whether they’re on-network or off-network, and that’s one use case that led us down this path of investing in endpoint protection.
Another one is that we see a lot of highly targeted attacks that are utilizing a threat that’s never been seen before and has been designed in such a way that it’s able to evade detection at the network security level. It could be based on a new zero-day vulnerability the attacker will use against a high-value target. Because this is based on an unknown vulnerability it’s missed by IPS/IDS. Our approach is effective at learning from these new attacks and routing new defenses back to the infrastructure so if that type of threat is used again it will be blocked. But if the attacker only uses it once then other areas of defense must kick in to protect an organization.
So those use cases are why we made the investment in Cyvera, and the release of Traps is our first official release of this technology and includes some integration into WildFire, which is our sandboxing technology.
The classic endpoint protection companies that offer antivirus-based protection rely on signatures for defense, which requires prior knowledge of the threat in order to block it. So these vendors have large teams of people who are constantly churning out signatures based on new threats they observe in the wild.
The challenge we saw with that approach is you’re always several steps behind the attacker community. There’s literally millions of forms of new malware that get generated each year. On a daily basis we see an average of over 20,000 new forms of malware. So companies with AV-based solutions have to build signatures against all of those new forms, then distribute those signatures out to all the endpoints. It’s an impossible situation to stay on top of.
Similarly, technologies like discreet intrusion prevention or intrusion detection systems require prior knowledge to protect against vulnerabilities. So if it’s an unknown zero-day based vulnerability, IPS or IDS isn’t as effective. It can only block what it knows.
So when we were looking at making an investment we spent a lot of time in our due diligence looking at the approaches that others use. There are a lot of companies jockeying for the space, knowing the traditional approaches are ineffective.
And we saw two common approaches we didn’t like as far as the new technology goes. The first was container-based tools that are basically designed to wrap a protective barrier around processes so if the process turns out to be malicious in nature the container detects it and shuts it down. But a lot of attackers have figured out how to disable those containers, and they impose a significant amount of resource overhead. So from an efficacy and operational perspective it wasn’t a very viable option.
Then the other approach that concerned us was tools focused on post-attack detection or remediation. You would deploy those to try and identify and isolate systems that were affected and then begin the cleanup process. If people are investing in that as their answer to highly targeted attacks, then they’re effectively waving a white flag, saying I can’t prevent these attacks so I might as well invest money in trying to at least detect them quickly.
We vehemently disagree with that premise. We do think that attacks, no matter how sophisticated, can be prevented. There is no silver bullet in this battle but network security will absolutely continue to play a big role in preventing attacks. But there are some holes that you have to shore up and that’s why we brought Traps to market.
Traps is a technology that, thus far, with the trials that we’ve done with different customers, has proven to be 100% effective against even the most highly targeted, zero-day based attacks.
How does it work?
What we liked about the technology is it’s not focused on the individual threat. Traps really doesn’t care whether it’s known or unknown malware. Traps doesn’t really care about the vulnerability itself. What Traps focuses on is the underlying techniques that an attacker must execute in order to exploit a vulnerability on an endpoint.
Let’s say an attacker found some sort of weakness in a piece of software and intended to use that to exploit the system. The attacker would have to go through a series of well-defined steps to make that happen. It may be three steps, it may be five steps. It depends on the nature of the exploit, but they would have to go through a sequence of steps. With Traps, what we’ve done is built a series blocks against each and every one of those available techniques so the second an attacker tries to employ one they run into a block and their attack is thwarted and the process is shut down. Today there are around two dozen techniques at an attackers disposal.
So let’s say there was a weakness in an Adobe PDF file and someone has initiated an exploit to try and take advantage of that weakness. As they go through the steps of that exploit, they would run into one of our exploit prevention modules within Traps and, as soon as they do, our product will shut down that process and alert the user that an attack was prevented and then also alert the admin. Then we collect a package of forensics, including memory state, etc., and provide it to the admin so they know the details of the attack, what user they were going after, what file they were using, etc.
And it is client based?
Right. Traps is a very thin client that lives on the endpoint itself. One of our criteria was this couldn’t be some big, heavy, resource-intensive type of technology. It literally consumes only 5MB of memory and about a tenth of one percent on average of CPU utilization. And it basically sits on that endpoint and anytime a new process is opened we inject what we call prevention modules into that process. So the second an attacker tries to utilize one of these known techniques they will run into one of our prevention modules and the attack is prevented.
How can you possibly account for all the different approaches that a vulnerability exploit would attempt?
Right now there are a total of 24 techniques that attackers have at their disposal to try and exploit a system, so we have that covered. These techniques are pretty hard science. It’s rare if you see two or three new techniques emerge within a year’s period of time. In fact, in the release that we announced we added three new prevention modules against three new techniques that emerged and those are the first techniques that we’ve seen in two years.
The vast majority of the techniques come out of academia. Someone in academia will be studying different processes, then publish a paper and attackers get a hold of that and, voila, they’ve got a new technique at their disposal. So we’ve been working very closely with academia to make sure that, as these things are being researched, we’re also building prevention modules against them so that when they publish their paper we also have modules built against those new techniques.
I suspect it will probably be another eight to twelve months or so before we see another one of these techniques emerge. They don’t happen that often.
I presume the tool is operating system dependent.
Correct. We support Windows XP, Windows 7 and Windows 8 on the workstation side, and on the server side it’s Windows Server 2003, 2008 and 2012. It sits well below the application stack so it’s independent of the applications themselves. So we support any kind of application that works on top of a Microsoft Windows environment.
In fact, I was talking to an oil and gas company and, while the prevention characteristics of this are very enticing, this guy was excited about the fact we support XP because he had tens of thousands of systems that were still running Windows XP and Microsoft isn’t patching XP anymore. So he was looking at this as a way to extend the lifespan of his Windows XP systems, which is a nice aftereffect. We’re seeing Windows in ATMs, point-of-sale systems, etc.
So that’s the exploit side, what about malware-based attacks?
Right. On the malware side it works similar, only we’ve added a couple of other steps. When it comes to malware-based attacks the process is slightly different. Malware of course doesn’t require a vulnerability exploit in order to run on an endpoint. Often it’s our employees who initiate this process by opening a malicious file attachment in email, clicking on a link that takes that person to a malicious URL or domain, downloading a malicious file from a USB stick, etc.
Traps malware prevention is accomplished in three steps. First, Traps allows admins to create a series of policies on the endpoint that significantly limits the risk of employees inadvertently downloading malware. These are simple policies like – do not allow a user to execute a .exe file sent over email, or from a removable storage device. By establishing the correct policies up front an organizations can reduce the options available for an attacker to get malware to an endpoint.
Second, Traps integrates with WildFire to provide an immediate vehicle to verify whether a file is known to be malicious. Every day WildFire inspects millions of files for new forms of malware. This intelligence is made available to Traps so it can verify whether a particular executable is malicious before allowing it to run on an endpoint. And finally, Traps utilizes malware prevention modules on the endpoint to ensure that the malware never executes.
Are competitors doing anything similar?
The only other company who’s kind of taken this approach is Microsoft themselves. There’s a project that Microsoft had been playing with called EMET and they’re the only ones really today that are focused on a technique-based approach. Microsoft has chosen not to productize EMET, but it’s kind of a skunksworks project, if you will. So really only us and Microsoft are the two that are looking at this from a techniques basis. And the EMET project only supports seven exploit techniques today.
What percentage of the problem do you think this addresses? After all, there’s environments other than Windows and there’s the whole mobility threat. How do you add that up?
Today Traps is focused on Windows-based support which constitutes the majority of endpoints. We plan to expand support in the future based on customer needs.
How do you sell this?
It is sold as a subscription service. So you can buy Traps as a one, three or five-year subscription and, as I mentioned, there is a thin client you have to deploy. It can be deployed through a company’s standard distribution software.
So a per-device fee?
Right now we have two price points, one for workstation and one for server. Then it’s on a tiered structure, with different price bands depending on the total number of deployed endpoints.
One more thing I want to mention. You’ll see us referring to Advanced Endpoint Protection, which we’re defining differently than how others might define endpoint protection today. Many definitions largely align with classic anti-virus capabilities. We think to qualify as an Advanced Endpoint Protection solution you have to be able to block all exploits, whether they’re known or unknown. You have to be able to block all malware, both known and unknown. Forensics remains crucial because there’s knowledge and insight that can be gained to protect the rest of the organization. It has to be very scalable and lightweight. If you’re deploying hundreds of thousands of these clients across endpoints as small as a point-of-sale system, this can’t be a big memory and CPU hog.
And finally, it has to be integrated with the cloud and the network. These worlds are going to collide in a very big way. If you can link the network with the endpoint and the endpoint with the network, there is a tremendous advantage across both fronts when it comes to ultimately bolstering security efficacy. They’re going to see things inherently the others can’t see, and if you can bring that together in terms of some type of sharing relationship, then everything becomes strong together.