We propose a study to assess the feasibility and efficacy of using graph/based analytics and Graph Neural Networks (GNN) to identify anomalous individual and organizational personas in financial transaction datasets. Currently, there is an urgent and expensive need to deny nefarious transnational state and non/state actors from accessing global financial systems, export/controlled technologies, critical supply chain networks, and other sensitive systems pertinent to U.S. national security. While various techniques exist today that perform pattern recognition and anomaly detection reporting, these techniques often result in limited operational value for key stakeholders and decision makers. As such, our analysis intends to quantify the impact of automated graph/based anomaly detection techniques for providing timely, relevant, actionable reporting in the financial threat intelligence domain. This Phase I Small Business Innovation Research (SBIR) proposal presents our research plan that aims to assess these objectives to determine future courses of action and best practices to achieve automated graph/based generative/alert reporting and analyst/defined discovery in databases. Finally, our research intends to provide a Technical Design Document (TDD) that holistically examines how graph/based techniques may be applied across SOCOMs net/new (Greenfield) and legacy modified (Brownfield) systems.
