Automated extraction of polymorphic virus signatures using abstract interpretation

In this paper, we present a novel approach for the detection and signature extraction for a subclass of polymorphic computer viruses. Our detection scheme offers 0 false negative and a very low false positives detection rate. We use context-free grammars as viral signatures, and design a process able to extract this signature from a single sample of a virus. Signature extraction is achieved through a light manual information gathering process, followed by an automatic static analysis of the binary code of the virus mutation engine.