Now You See Me: Identifying Duplicate Network Personas

This work provides a decision-making framework at the intersection of social network analysis and law enforcement intelligence with the goal of identifying persons of interest in a social network. Criminal social networks are complex due to the limited and imperfect information available. Moreover, the participating entities tend to misrepresent themselves in order to stay hidden and covert. In this work, we propose a new integer programming formulation to assist in the identification of entities who are prone to misrepresent themselves in a social network. Our insight is that such personas will form large subgraphs of restricted diameter that are connected to other entities who do not communicate directly or within a short number of intermediates. We formally define the problem and derive its computational complexity. Additionally, we provide an integer programming formulation to solve it exactly with the use of a commercial solver. We then show how our framework behaves on the Krebs 9/11 network. Our approach is able to identify what are believed to be two distinct clusters of criminals participating in two separate subplots: the multiple flight hijacking on September 11; as well as a plot against the U.S. embassy in Paris in the year 2001.