Detecting Deception Within a Probabilistic Modelling Framework

In this paper, we consider a number of analytic approaches to identifying or accounting for possible deception tactics being employed by an adversary. These are equally applicable to military or civil intelligence, or even law enforcement. As well as examining the Analysis of Competing Hypotheses (ACH) methodology, employed by some intelligence agencies, we focus on the contribution of methods for reasoning under uncertainty, particularly Bayesian networks (BNs). We also discuss the combination of these approaches as suggested by other authors. It is shown that the incorporation of negative evidence in addition to positive observations improves the performance of the BNs.