Development Of Novel Approaches To Anomaly Detection And Surety For Safeguards Data - Year One Results

Year
2020
Author(s)
Natacha Peter Stein - Sandia National Laboratories
David Farley - Sandia National Laboratories
Constantin Brif - Sandia National Laboratories
Nicholas Pattengale - Sandia National Laboratories
Chase Zimmerman - Sandia National Laboratories
Yifeng Gao - George Mason University
Jessica Lin - George Mason University
Mitchell Negus - University of California, Berkeley
Rachel Slaybaugh - University of California, Berkeley
Meghan Galiardi - Sandia National Laboratories
Abstract

The Novel Approaches to Anomaly Detection and Surety for Safeguards Data project, which was introduced at the Institute for Nuclear Materials Management annual meeting in 2019, investigates three core data analysis and management methods and their applicability for international safeguards: Distributed Ledger Technology (DLT) for data authentication, anomaly detection based on Grammar Compression (GC), and how operator data could assist in drawing safeguards conclusions in a Multi-Party Computing (MPC) environment. This paper outlines the work performed in Year one of the project, highlighting results and their impact on the continuation of the tasks. For DLT technologies, an experimental comparison of current safeguards practice versus a DLT-backed prototype is described. For GC-based anomaly detection, we present the investigation of new methods with improved performance based on ensemble learning and variable-length motif discovery. With regards to MPC-based data protection, the viability of applying the method for anomaly detection is analyzed. These three main initiatives are being developed in parallel, with constant cross-effort interactions, and the joint goal of creating an integrated demonstration platform that combines a DLT prototype for safeguards data authentication, GC-based anomaly detection, and MPC-based data integration for sensitive facility information. An outlook on the remaining work for Year two of the project concludes the paper. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525. SAND No. 2020-1444.