Model-based intrusion detection by abstract interpretation

Model-based intrusion detection works by comparing a process's runtime behavior with a pre-computed normal program model. This paper studies this technology from the viewpoint of abstract interpretation theory. We regard different program behavior models used to perform intrusion detection as different abstractions of the concrete trace semantics of programs. Based on this point, we formally define model-based intrusion detection and present a generic generation algorithm for program models on a provided abstraction domain. Eventually, we discuss how to use this mechanism to implement a real intrusion detection model proposed by us before.