In a panel discussion I had at EIC 2014 with Roy Adar, Vice President of Product Management at CyberArk, Roy brought up an interesting number: according to research, attacks start on average 200 days before they are detected. Taking into account the Gaussian distribution behind this average, some attackers might have been active for years before they were detected. And who knows whether all of them are detected at all.
How to react to this? There are several elements in the answer. Protect your systems with various layers of security. Use anti-malware tools, even while they won’t catch every malware and every attacker. Encrypt your sensitive information. Educate your employees. These and other “standard” actions are quite common. But there is at least one other thing you should do: analyze the behavior of users in your network.
I do not mean user tracking in the sense of “do they do their job” (which is hard to implement in countries with strong worker councils), I’m talking about identifying anomalies in their behavior. Attackers are characterized by uncommon behavior. Users might access far more documents than average or than they did before. Accounts might be used at unusual times. Users might log in from suspicious locations. Sometimes, it is not a single incident, but a combination of things, eventually over a longer period of time, which is typical for a specific form of attack, especially in the case of long-running APTs (Advanced Persistent Threats).
There is an increasing number of technologies available to analyze such patterns. Standard SIEM (Security Information and Event Management) tools are one approach, however analysis of anomalies might be difficult to perform based on rules. However, there is an increasing number of solutions that rely on more advanced pattern-matching technologies. These can, based on specific mathematical algorithms, turn log events and other information into patterns (in fact complex matrices), and analyze these for anomalies. There might be some noise in the sense of false negatives in the results, but this is true for rule-based analytics as well. Combination of such analytical technologies can make a lot of sense – if you bring together specialized analytics for areas such as Privilege Management (for instance, CyberArk’s PTA), User Activity Monitoring, pattern-based analytics, and traditional SIEM, you might learn a lot about these anomalies and hence about the attacks that are already running and the attackers behind them.
Even while you might feel it being too early to move towards RSI, you should put your focus on how to learn more about the attackers that are already inside your network. Understanding anomalies and patterns with new types of analytical technologies might help.