This paper presents a new approach to run-time security monitoring that can detect system abnormalities including attacks,
faults, or operational errors. The approach, System Health and Intrusion Monitoring (SHIM), employs a hierarchy of constraints
to describe correct operation of a system at various levels of abstraction. The constraints capture static behavior, dynamic
behavior, and time-critical behavior of a system. A system in execution will be monitored for violation of the constraints,
which may indicate potential security problems in the system. SHIM is based on specification-based intrusion detection, but
it attempts to provide a systematic framework for developing the specifications/constraints. SHIM does not detect directly
the intrusive actions in an attack, but their manifestations as violations of constraints. In this paper, we describe the
constraint model and the methodology for developing the constraints. In addition, we present preliminary results on the constraints
developed for host programs and network protocols. By bounding the behavior of various system components at different levels
of abstraction, SHIM has a high chance of detecting different types of attacks and their variants.
This research is supported by Defense Advanced Research Project Agency (DARPA) under contract F30602-00-C-0210.