Detecting Suspicious Behavior from Only Positional Data with Distributed Sensor Networks

Abstract

Wireless sensor networks are increasingly popular, and are being used to measure simple properties of their environment. In many applications such as surveillance, we would like them to distinguish "suspicious" behavior automatically. We distinguish here between suspicious and anomalous behavior, and develop a mathematical model which we illustrate on some sample data. We show the model predicts six classic deception strategies. We conclude with analysis of more sophisticated deceptions that exploit system responses to simpler deceptions.