Malware analysis with graph kernels and support vector machines

This paper addresses a fundamentally new method for analyzing the behavior of executed applications and sessions. We describe a modeling framework capable of representing relationships among processes belonging to the same session in an integrated way, as well as the information related to the underlying system calls executed. We leverage for this purpose graph-based kernels and support vector machines (SVM) in order to classify either individually monitored applications or more comprehensive user sessions. Our approach can serve both as a host-level intrusion detection and application level monitoring and as an adaptive jail framework.