Development of Novel Approaches to Anomaly Detection and Surety for Safeguards Data

The verification tasks for international nuclear safeguards rely heavily on data coming from International Atomic Energy Agency (IAEA) safeguards equipment installed at facilities. Within the current verification paradigm, this entails using expensive data authentication and encryption, which (in conjunction with operator requests/preferences) can limit the broad deployment of IAEA safeguards equipment on-site. In this project we will evaluate a new safeguards data paradigm that will develop and test a novel safeguards data authentication, integration, and analysis workflow on the foundation of distributed ledger technology (DLT) and anomaly detection based on the grammar compression (GC) method. For data authentication, we will design and prototype a pilot safeguards data tracking system that will take advantage of recent advances in provenance tracking based on DLT. While GC method will provide practical method for effective and efficient detection of anomalies in multivariate time-series data obtained from fielded safeguards-like equipment. We will also explore the potential use of operator data such as safety and physical protection systems’ data to augment traditional safeguards verification data within secure, multi-party computing (MPC) environments that would protect an operator’s commercial proprietary or otherwise sensitive data while allowing a safeguards inspectorate to gain confidence that anomalous activity within a facility could be attributed to safety or security activities rather than diversion or misuse at the facility. These two approaches are natural complements to one another and can find applications in both domestic and international safeguards. The DLT and GC-based anomaly detection can be used on traditional safeguards data sources or other available data, and the multi-party computing component allows for exploration of nontraditional data sources in a manner that will protect sensitive operator information. For the development and testing of these approaches we will utilize data collected from the Multiple-Informatics for Nuclear Operations Scenarios (MINOS) venture testbed as a proxy for international safeguards data. This paper outlines the work that is planned over the full duration of the project, potential impact of the anticipated results, as well as preliminary results of the work performed up to date.