New perspectives in criminal network analysis: multilayer networks, time evolution, and visualization

The work presented in this Dissertation reflects a long-term human, professional and cultural path started some years ago when I first developed LogAnalysis, a tool for the analysis and visualization of criminal and social networks. Since then, I devoted myself to the development of frameworks, algorithms and techniques for supporting intelligence and law enforcement agencies in the task of unveiling the CN structure hidden in communication data, identifying the target offenders for their removal or selecting effective strategies to disrupt a criminal organization. In a natural way, I successively focused on the evaluation of the resilience of criminal networks and on the multiplex formalism, which takes into account the various relationships existing within a criminal organization. In this context I introduce criminal network analysis tools: LogAnalysis, LogViewer, Semantic viewer and Failure simulator. I have been involved in the design, modeling, and writing of all of the works presented. In particular, I have also developed and tested all the visual tools included therein. Finally, I introduce Multiplex PBFS (Mx-PBFS) a novel multi-threaded parallel Breadth-First Search algorithm for categorical and inter-layer couplings multiplex networks, and the framework CriMuxnet (still under development) for multilayer criminal networks analysis based on high-quality 3D visualizations of network data. CriMuxnet was designed to work in conjunction with a 3D computer graphics (CG) packages: Autodesk Maya or Blender. CriMuxnet exploits 3D engine features to significantly improve both exploratory search and visualization strategy.