SRI Security Research Institute, Edith Cowan University, Perth, Western Australia
Assessing the credibility of an evidential source is an important part of intelligence analysis, particularly where human intelligence is concerned. Furthermore, it is frequently necessary to combine multiple items of evidence with varying source credibilities. Bayesian networks provide a powerful probabilistic approach to the fusion of information and are increasingly being applied in a wide variety of settings. In this paper we explore their application to intelligence analysis and provide a simple example concerning a potential attack on an infrastructure target. Our main focus is on the representation of source credibility. While we do not advocate the routine use of quantitative Bayesian networks for intelligence analysis, we do believe that their qualitative structure offers a useful framework for evidence marshalling. Furthermore, we believe that quantified Bayesian networks can also play a part in providing auxiliary models to explore particular situations within a coherent probabilistic framework. This process can generate fresh insights and help to stimulate new hypotheses and avenues of enquiry.