Runtime detection of zero-day vulnerability exploits in contemporary software systems

Abstract

It is argued that runtime verification techniques can be used to identify unknown application security vulnerabilities that are a consequence of unexpected execution paths in software. A methodology is proposed that can be used to build a model of expected application execution paths during the software development cycle. This model is used at runtime to detect exploitation of unknown security vulnerabilities using anomaly detection style techniques. The approach is evaluated by considering its effectiveness in identifying 19 vulnerabilities across 26 versions of Apache Struts over a 5 year period.