Key points are not available for this paper at this time.
One of the primary challenges in intrusion detection is modelling typical application behavior so that we can recognize attacks by their atypical effects without raising too many false alarms. We show how static analysis may be used to automatically derive a model of application behavior. The result is a host-based intrusion detection system with three advantages: a high degree of automation, protection against a broad class of attacks based on corrupted code, and the elimination of false alarms. We report on our experience with a prototype implementation of this technique.
Wagner et al. (Wed,) studied this question.