Analysing The Uncertainty of Wiretaps in Criminal Networks #

Daniel Catlin

11:50 Wednesday in 2Q49.

Part of the Networks and complex systems in society session.

Abstract #

Criminal investigations involving wiretaps are often long, expensive and conducted with very little information of the entire criminal organisation being investigated. Once a few members of the organisation are identified, surveillance can begin, and further co-conspirators identified with the aim to find the key players in the network. But how do we know who are the important persons in a network which is usually very decentralised, and how can we be sure of our findings to progress the investigation? We look at the communications data of criminal organisation as a time-dependent network of symmetric graphs and use the time evolving Katz network centrality to determine the most important nodes in the network based on the differing amount of information flow in the network and their broadcasting or receiving of it. From here we modify the network by introducing Bernoulli variables to encapsulate conversations that could exist but were not heard by law enforcement, using this we can create new node rankings of importance based on the likelihood of missing conversations for individuals under increased surveillance. From this we can then explicitly calculate the expectation and variance of the enhanced network matrices and study the stability of its values under small perturbations. We conduct this investigation on synthetically generated network data with differing characteristics and present results using a real-life case where data was collected over a period of 4 years. By analysing the stability of these systems, we can improve the efficacy and efficiency of these investigations to allow law enforcement to better allocate resources and come to more favourable conclusions.