Detecting Suspicious Behavior from Positional Information
Abstract
Suspiciousness is not the same an anomalousness. Suspicion requires evidence of deception in observed attempts at
concealment. We propose metrics for measuring suspiciousness of agents moving in a sensor field based on only periodic
knowledge of their positions (as with large numbers of "small and cheap" sensors). This has applications to electronic
sentries and counterterrorism. This theory requires assessment of the behavior, visibility, and noticeability of the average
agent as well as the anomalousness of the position, velocity, and acceleration vectors of a particular agent. We conclude with
a report on experiments with an implementation of our theory on a simulated sensor network.
Description
This paper appeared in the Proceedings of the Workshop on Modeling Others from Observations at the International Joint
Conference on Artificial Intelligence, Edinburgh, Scotland, July 2005.