Applied Network Security Analysis: The Breach Confirmation Use Case
As our last use case in Applied Network Security Analysis, it’s time to consider breach confirmation: confirming and investigating a breach that has already happened. There are clear similarities to the forensics use case, but breach confirmation takes forensic analysis to the next level: you need to learn the extent of the breach, determining exactly what was taken and from where. So let’s revisit our Forensics scenario to look at how that can be extended to confirm a breach. In that scenario, a friend at the FBI gave you a ring to let you know they found some of your organization’s private data during a cybercrime investigation. In the initial analysis you found the compromised devices and the attack paths, as part of isolating the attack and containing the damage. You cleaned the devices and configured IPS rules to ensure those particular attacks will be blocked in the future. You also added some rules to the network security analysis platform to ensure you are watching for those attack traffic patterns moving forward – just in case the attackers evade the IPS. But you still can’t answer the question: “What was stolen?” with any precision. We know the attackers compromised a device and used it to pivot within your environment to get to the target: a database of sensitive information. We can’t assume that was the only target, and if your attack was like any other involving a persistent attacker, they found multiple entry points and have a presence on multiple devices. So it’s time to head back into your analysis console to figure out a few more things. What other devices were impacted: Start by figuring out how many other machines were compromised by the perimeter device. By searching all traffic originating from the compromised DMZ server, you can see which devices were scanned and possibly owned. Then you can confirm using either the configuration data you’ve been collecting, or by analyzing the machine using an endpoint forensics tool. You may already have some or all of this information already when trying to isolate the attack path. What was taken: Next you need to figure out what was taken. You already know at least one egress path, identified during the initial forensic analysis. Now you need to dig deeper into the egress captures to see if there were any other connections or file transfers to unauthorized sites. The attackers continue to improve their exfiltration techniques, so it’s likely they’ll use both encrypted protocols and encrypted files to make it hard to figure out what was actually stolen. Having the full packet stream allows you to analyze the actual files, though depending on the sophistication of your attacker, you might need specialized help (from third party experts or law enforcement) to break the crypto. Remember that the first wave of forensic investigation focuses on determining the attack paths and gathering enough information to do an initial damage assessment and remediation. From there it’s all about containing the immediate damage as best you can. This next level of forensic analysis is more comprehensive: determine the true extent of the compromise and inventory what was taken. As you can imagine, without having the network packet capture it’s impossible to do this analysis. You would be stuck with log files telling you what happened, but not what, how, or how much was taken. That’s why we keep harping on the need for more comprehensive data on which to base network security analysis. Clearly you can’t capture all the data flowing around your networks, so it’s likely you’ll miss something. But you will have a lot more useful information for responding to attacks than organizations which do not capture traffic. Summary To wrap up our Applied Network Security Analysis series let’s revisit some of the key concepts we’ve covered. First of all, in today’s environment, you can’t assume you will be able to stop a targeted attacker. It is much smarter and more realistic to assume your devices are compromised, and to act accordingly. This assumption puts a premium on detecting successful attacks, preventing breaches, and containing damage. All those functions require data collection to be able to understand what is happening in your environment. Log Data Is Not Enough: Most organizations start by collecting their logs, which is clearly necessary for compliance purposes. But it’s not sufficient – not by a long shot. Additional data – including configuration, network flow, and even the full network packet stream, is key to understanding what is happening in your environment. Forensics: We walked through how these additional data sources can be instrumental in a forensic investigation, ultimately resulting in a breach confirmation (or not). The key objective of forensics is to figure out what happened, and the ability to replay attacks and monitor egress points (using the actual traffic) replaces forensic interpretation with tested fact. And forensics folks like facts much better. Security: These additional data sources are not just useful after an attack has happened, but also to recognize issues earlier. By analyzing the traffic in your perimeter and critical network segments, you can spot anomalous behavior and investigate preemptively. To be fair, the security use cases are predicated on knowledge of what to look for, which is never perfect. But in light of the number of less sophisticated attackers using known attacks, making sure you don’t get hit by the same attack twice is very useful. We all know there are always more projects than your finite resources and funding will allow. So why is network security analysis something that should bubble up to the top of your list? The answer is about what we don’t know – we cannot be sure what the attack will be in the future, but we know it will be hard to detect, and it will steal critical information. We built the Securosis security philosophy on Reacting Faster and Better, by focusing on gathering as much data as possible, which requires an institutional commitment to data collection and analysis.
