Skip to main content
SHARE
Project

Situ

Project Details

Principal Investigator
Funding Source
Laboratory Directed Research and Development (LDRD)
Start Date
Situ Streaming Visualization

Situ: Discovering Suspicious Behavior in Cyber Security

Despite the best efforts of cyber security analysts, networked computing assets are regularly compromised, resulting in the loss of intellectual property, the disclosure of state secrets, and financial damages in the billions. A 2014 report from the Center for Strategic and International Studies estimated the global cost of cyber crime at $400 billion annually. There has also been a rise of sophisticated attack groups that continually develop novel methods of penetrating networks that current technologies are typically unable to detect. Signature-based security systems are effective at detecting known attacks, but are unable to detect novel or sophisticated attacks. Indeed, automated security systems will never be capable of detecting all malicious activity. Network operators need tools to help identify suspicious behavior that bypasses automated security systems.

Situ combines anomaly detection and data visualization to provide a distributed, streaming platform for discovery and explanation of suspicious behavior to enhance situation awareness. Our novel approach to anomaly detection is based on unsupervised, probabilistic modeling. Key to our approach is modeling events in different contexts or at multiple scales; each event is modeled and scored by multiple anomaly detectors to identify different kinds of anomalous behavior. The architecture of Situ is designed to scale to very high data rates on commodity hardware—hundreds of thousands of events per second.

Situ helps network operators discover and understand suspicious events that would otherwise go undetected. It reduces the huge volumes of raw network data to a smaller, manageable number of events that should be examined by human domain experts. By highlighting suspicious activity operators can find novel attacks, but can also be made aware of insider threats, policy violations, misconfigurations, and new kinds of behavior that may require some investigation. Through the application of multiple contexts, Situ can look for a wide range of activity. Different contexts perform better for different kinds of attacks. Multiple contexts can also help explain why an event is suspicious since the varying scores will point operators at certain kinds of behaviors.

 

Contact

Senior Researcher, Data Science and Visualization Group
John R Goodall