As we discussed in our last post, there is a logical lifecycle which you can implement to protect endpoints. Once you know what you need to protect and how vulnerable the devices are, you try to prevent attacks, right? Was that a snicker? You’ve been reading the trade press and security marketing telling you prevention is futile, so you’re a bit skeptical. You have every right to be – time and again you have had to clean up ransomware attacks (hopefully before they encrypt entire file servers), and you detect command and control traffic indicating popped devices frequently. A sense of futility regarding actually preventing compromise is all too common.
Despite any feelings of futility, we still see prevention as key to any Endpoint Protection strategy. It needs to be. Imagine how busy (and frustrated) you’d be if you stopped trying to prevent attacks, and just left a bunch of unpatched Internet-accessible Windows XP devices on your network, figuring you’d just detect and clean up every compromise after the fact. That’s about as silly as basing your plans on stopping every attack.
So the key objective of any prevention strategy must be making sure you aren’t the path of least resistance. That entails two concepts: reducing attack surface, and risk-based prevention. Shame on us if devices are compromised by attacks which have been out there for months. Really. So ensuring proper device hygiene on endpoints is job one. Then it’s a question of deciding which controls are appropriate for each specific employee (or more likely, group of employees). There are plenty of alternatives to block malware attacks, some more effective than others. But unfortunately the most effective controls are also highly disruptive to users. So you need to balance inconvenience against risk to determine which makes the most sense. If you want to keep your job, that is.
“Legacy” Prevention Techniques
It is often said that you can never turn off a security control. You see the truth in that adage when you look at the technologies used to protect endpoints today. We carry around (and pay for) historical technologies and techniques, largely regardless of effectiveness, and that complicates actually defending against the attacks we see.
The good news is that many organizations use an endpoint protection suite, which over time mitigates the less effective tactics. At least in concept. But we cannot fully cover prevention tactics without mentioning legacy technologies. These techniques are still in use, but largely under the covers of whichever endpoint suite you select.
- Signatures (LOL): Signature-based controls are all about maintaining a huge blacklist of known malicious files to prevent from executing. Free AV products currently on the market typically only use this strategy, but the broader commercial endpoint protection suites have been supplementing traditional signature engines with additional heuristics and cloud-based file reputation for years. So this technique is used primarily to detect known commodity attacks representing the low bar of attacks seen in the wild.
- Advanced Heuristics: Endpoint detection needed to evolve beyond what a file looks like (hash matching), paying much more attention to what malware does. The issue with early heuristics was having enough context to know whether an executable was taking a legitimate action. Malicious actions were defined generically for each device based on operating system characteristics, so false positives (notably blocking a legitimate action) and false negatives (failing to block an attack) were both common – a lose/lose scenario. Fortunately heuristics have evolved to recognize normal application behavior. This dramatically improved accuracy by building and matching against application-specific rules. But this requires understanding all legitimate functions within a constrained universe of frequently targeted applications, and developing a detailed profile of each covered application. Any unapproved application action is blocked. Vendors need a positive security model for each application – a tremendous amount of work. This technique provides the basis for many of the advanced protection technologies emerging today.
- AWL: Application White Listing entails implementing a default deny posture on endpoint devices (often servers). The process is straightforward: Define a set of authorized executables that can run on a device, and block everything else. With a strong policy in place, AWL provides true device lockdown – no executables (either malicious or legitimate) can execute without explicit authorization. But the impact to user experience is often unacceptable, so this technology is mostly restricted to very specific use cases, such as servers and fixed-function kiosks, which shouldn’t run general-purpose applications.
- Isolation: A few years ago the concept of running apps in a “walled garden” or sandbox on each device came into vogue. This technique enables us to shield the rest of a device from a compromised application, greatly reducing the risk posed by malware. Like AWL, this technology continues to find success in particular niches and use cases, rather than as a general answer for endpoint prevention.
You can’t ignore old-school techniques, because a lot of commodity malware still in circulation every day can be stopped by signatures and advanced heuristics. Maybe it’s 40%. Maybe it’s 60%. Regardless, it’s not enough to fully protect endpoints. So endpoint security innovation has focused on advanced prevention and detection, and also on optimizing for prevalent attacks such as ransomware.
Let’s unpack the new techniques to make sense of all the security marketing hyperbole getting thrown around. You know, the calls you get and emails flooding your inbox, telling you how these shiny new products can stop zero-day attacks with no false positives and insignificant employee disruption. But we don’t know of any foolproof tools or techniques, so we will focus the latter half of this series on detection and investigation. But in fairness, advanced techniques do dramatically increase the ability of endpoints to block attacks.
The first major category of advanced prevention techniques focus on blocking exploits before the device is compromised. Security research has revealed a lot of how malware actually compromises endpoints at a low level, so tools now look for those indicators. You can pull out our favorite healthcare analogy: by understanding the fundamental changes an attack causes within an organism, you learn what to look for generally, rather than focusing on a specific attack, which can morph in an infinite number of ways.
These tactics break down into a few buckets:
- Profiling exploit behavior: This takes the advanced heuristics approach described above deeper into the innards of the operating system. Where the advanced heuristics focus on identifying anomalous application behavior, these anti-exploit tools focus on what happens to the actual machine when malicious code takes over the device. The concept is that there are a discrete and known number of ways to compromise the operating system, regardless of the attack vector and by blocking those behaviors, you stop the exploit.
- Memory analysis/protection: One of the latest waves of attack doesn’t even deal with traditional malware files. Malicious code is inserted directly into a command line or other means of manipulating the operating system without hitting disk. This attack requires analyzing the memory of the device on a continuous basis and preventing memory corruption and logic flaws. Suffice it to say this kind of technology is very sophisticated and can really impact the operation of the device, so full testing to ensure no impact on your devices is critical in evaluating this technology.
- Malware-less defense: Aside from hiding attacks in memory, attackers are now using fundamental operating system features to defeat whitelisting and isolation techniques. The most frequently targeted OS services include WMI, PowerShell, and EMET. These attacks are much more challenging to detect because these system processes are authorized by definition. To defend against these attacks, advanced technologies need to monitor the behaviors of all processes to make sure an approved process hasn’t been hijacked. This requires profiling legitimate behavior of common system processes, and then looking for anomalous activity.
All ‘advanced’ endpoint protection technology includes these techniques, though they may be branded differently. It is all largely the same approach of looking for anomalous behavior, but focused on OS and device innards instead of user-space applications.
Endpoint Bot Detection
Pretty much every modern attack, whether it involves malware or not, involves communicating with a command and control network to download the attack payload and receive instructions. So endpoint network-based detection has evolved to look for command and control patterns, similar to non-endpoint network malware detection.
This capability is important for full protection, because endpoints aren’t always on the corporate network, which you are presumably already scanning for command and control traffic. So recognizing when a device in a coffee shop or hotel is communicating with known malicious sites, can help you detect a compromise before the device reconnects to the corporate network. This requires integration with a threat intelligence source to keep an updated list of known malicious sites.
Dynamic File Testing
Many attacks still involve a compromised file executing code on a device, so network and cloud sandboxes are heavily used to dynamically execute inbound files and ensure they are not malicious. You have a number of options for where to test files, including the perimeter and/or email security gateway. But remote personnel remain a challenge because their network traffic doesn’t run through the network’s corporate defenses.
So you can supplement those corporate controls with the ability to extract and test files on endpoints as well. The file will be checked to see if it has a known bad hash; if not, it can be tested in the corporate sandbox. Some organizations now converting any easily compromised file (meaning Office files) into a sanitized PDF to remove any active code without impacting document appearance. If the original file is needed it can be routed to the recipient after clearing the sandbox.
The first technology you have certainly been hearing a lot about is machine learning. It is used in many contexts aside from endpoint protection, but the advanced endpoint security messaging has become very prominent. We just chuckle – statistical analysis of malware has been a popular technique as long as we can remember. And all of a sudden, math is our savior, to stop all these nasty attacks?
But the math really is better now. Combined with much more detailed understanding of how malware actually compromises devices, more sophisticated static file analysis does help detect attacks. But we have to wonder whether these new techniques are really just next-generation AV signatures.
Ultimately we try to avoid getting wrapped up in vernacular or semantics. If these techniques help detect attacks more accurately at scale, the important thing isn’t whether they look like signatures or not. It’s not like we (or anyone else) believe machine learning is the perfect solution for endpoint protection. It’s just another development in the never-ending arms race of malware protection.
The other enabling technology that warrants mention is threat intelligence. Or security research, as endpoint protection vendors have been calling it for a decade. The reality is that whether you are adding new indicators to an endpoint agent, or updating the list of known malicious sites for command and control detection, each endpoint agent needs to be updated frequently to keep current. Especially devices that don’t sit behind the corporate network’s perimeter defenses.
You wouldn’t necessarily buy threat intelligence as part of an endpoint protection project, but during technology evaluation you should ensure that agents are kept current, and updates don’t put too much strain on either endpoints or the network.
Protecting the Point of Attack
We should address the best place to place protection, because you have a few options. The path of least resistance remains network-based solutions, which can be deployed without any user impact. Of course these options don’t protect device which aren’t behind the corporate perimeter. Nor can network-based solutions provide context for individual user behavior like something running on the device can.
You can run all traffic through a VPN or a cloud-based filtering service to provide some protection for remote devices. Running traffic through either enables yoyu to gather telemetry and enforce corporate usage policies. On the downside, this impacts traffic flow and can be evaded by both savvy users and attackers. But it offers an option for addressing the limitations of filtering traffic through network defenses.
But this research is focused on endpoint protection, so let’s assume that protecting endpoints is important. So do you add yet another agent to your endpoint, or use a plug-in into a common application like a browser to protect against the most common attack vector? If for some reason you cannot replace the existing endpoint agent, looking at a plug-in approach to provide additional protection can certainly help as a stopgap.
But if we haven’t yet made it clear, these advanced endpoint security offerings are neither a long-term alternative, nor meant to run alongside an existing endpoint protection suite. These new offerings represent an evolution of endpoint protection; so either incumbents will add these capabilities to their existing offerings or they won’t survive. And this is not just about prevention – we will discuss endpoint detection and response capabilities in our next post.
We don’t normally call out specifically attacks because they change so frequently. But ransomware is a bit different. The ability to so cleanly and quickly monetize successful attacks has made it the most visible attack strategy. And ransomware is not restricted to just one size or type of company, or device type. We have seen ransomware targeting everyone and everything.
So how can you combine these advanced techniques to prevent a ransomware attack? Fortunately in technical terms ransomware is just another attack, so it can be profiled and blocked using advanced heuristics and exploit profiling. First look for attack patterns as they attempt to compromise the device; ransomware doesn’t look fundamentally different than other attacks.
Next look for clues within the endpoint’s network stack – particularly command and control traffic – because attackers need to deliver their payload to lock down the machine. You can also look for anomalous searching of file shares because ransomware typically targets shared file systems for extra impact.
Additionally, because ransomware encrypts the local file system, you can monitor file I/O for anomalous activity. We also suggest organizations more aggressively monitor their storage networks and arrays for anomalous file activity. This can help shorten the detection window, and stop encryption before too much data is impacted.
And yes, they are out of the scope for this research, but device and data backup are essential for quick restoration of service in case of a ransomware attack.
A Note on ‘Effectiveness’
It’s worth mentioning how to evaluate the effectiveness of these solutions. We refer back to our Advanced Endpoint and Server Protection research a few years ago, as this material hasn’t changed.
As you start evaluating these advanced prevention offerings, don’t be surprised to get a bunch of inconsistent data on the effectiveness of specific approaches. You are also likely to encounter many well-spoken evangelists spouting monumental amounts of hyperbole and religion in favor of their particular approach – whatever it may be – at the expense of all other options. This happens in every security market undergoing rapid innovation, as companies try to establish momentum for their approaches and products.
A lab test favoring one product or approach over another isn’t much consolation when you need to clean up an attack your tools failed to prevent. And those evangelists are nowhere to be found when a security researcher shows how to evade their shiny technology at the latest Black Hat conference. We at Securosis try to float above the hyperbole and propaganda to keep you focused on what’s really important – not claimed 1% effectiveness differences. If products or categories are within a few percent of each other across a variety of tests, we consider that a draw.
But if you look hard enough, you can find value in comparative tests. An outlier warrants investigation and a critical assessment of the test and methodology. Was it skewed toward one category? Was the test commissioned by a vendor or someone else with an agenda? Was real malware, freshly found in the wild, used in the test? All testing methodologies have issues and limitations – don’t base a decision, or even a short list, around a magic chart or a product review/test.
A Risk-Based Approach to Defending Endpoints
Yet, security practitioners have an unfortunate tendency to miss the forest for the trees when discussing advanced endpoint protection. The reality is that each device contains a mixture of data types; some data types present great risk to the organization, and others don’t. You also need to consider that some protection techniques are very disruptive to end users and can be expensive to both procure and manage.
So we advocate a risk-based approach to protecting endpoints. This involves grouping endpoint devices into a handful (or less than a handful) of risk categories. Then determine the most effective means to protect the devices based in each category. For example you might want to implement whitelisting on all kiosks in stores and warehouses. Or you might add an advanced exploit prevention agent to devices used by senior management, Human Resources, and Finance, and anyone else handling especially sensitive or attractive information. Finally you might just use free AV on devices which only have outbound access from common areas, because they don’t have access to anything important on the corporate network.
There are as many permutations as devices on your network. To scale this approach you need to categorize risk tiers effectively. But a one-size-fits-all approach doesn’t work either given the variety of different approaches that can be brought to bear on detecting advanced malware.
As we mentioned above, our next post will cover endpoint detection and response technologies which are increasingly important to defending endpoints.