Overview
Cross-jurisdictional law enforcement data sharing and analysis is of vital importance because law breakers regularly operate in multiple jurisdictions. Agencies continue to invest massive resources in various sharing initiatives despite several high-profile failures. Key difficulties include: privacy concerns, administrative issues, differences in data representation, and a need for better analysis tools. This work presents a methodology for sharing and analyzing investigation-relevant data and is potentially useful across large cross-jurisdictional data sets. The approach promises to allow crime analysts to use their time more effectively when creating link charts and performing similar analysis tasks. Many potential privacy and security pitfalls are avoided by reducing shared data requirements to labeled relationships between entities. Our importance flooding algorithm helps extract interesting networks of relationships from existing law enforcement records using user-controlled investigation heuristics, spreading activation, and path-based interestingness rules. In our experiments, several variations of the importance flooding approach outperformed relationship-weight-only methods in matching expert-selected associations. We find that accuracy in not substantially affected by reasonable variations in algorithm parameters and demonstrate that user feedback and additional, case-specific information can be usefully added to the computational model.
