Skip to main content
European Commission logo
ESARDA
Scientific paper

List Mode Inference Using Linear Classifiers for Nuclear Arms Control Verification

ESARDA Bulletin - The International Journal of Nuclear Safeguards and Non-Proliferation

Details

Identification
ISSN: 1977-5296, DOI: 10.3011/ESARDA.IJNSNP.2023.2
Publication date
19 May 2023
Author
Joint Research Centre

Description

Volume: 65, December 2023, pages 10-18

Authors: Eduardo Padilla1,2, Heidi Komkov2, Christopher Siefert2, Adam Hecht1, Ryan Kamm2, Kyle Weinfurther2, Jesus Valencia2

1University of New Mexico, Department of Nuclear Engineering, 2Sandia National Laboratories

Abstract: In potential future nuclear arms control treaties, methods to confirm the presence or absence of a nuclear warhead or nuclear components are likely to be a central function of a verification regime. Higher confidence in verification methods can be achieved through more rigorous, thus potentially sensitive, analysis of radiation signatures from treaty accountable items. Therefore, methods that protect sensitive information while allowing for rigorous analysis are a critical component of any potential nuclear treaty verification system; these methods are referred to as information barriers. In this paper, we describe the development of a novel radiation analysis method for listmode (time-stamped pulse heights, pulse-by-pulse) inference using linear classifiers trained on a large set of synthetically generated high resolution gamma spectra. In practice, each detector pulse would be fed into a linear classifier with the applied weight incrementing or decrementing counters for each class. After a set number of pulses, the highest output score determines the classification of the source of radiation. As such, this method serves as both a verification algorithm and information barrier combined. This new method achieves reliable discrimination (83% accuracy) of notional nuclear weapons grade treaty accountable item radiation signatures from those of a diverse, largely unconstratined, set of nuclear, medical and industrial radioisotope combinations. Importantly, this is shown to be achievable without the collection or processing of a potentially sensitive gamma radiation spectrum. This study serves as a proof of concept for the development of an intrinsic information barrier for attribute identification supporting nuclear arms control treaty verification.

Keywords: Information Barrier, Warhead Verification, Machine Learning

Reference guideline:

Padilla, E., Komkov, H., Siefert, C., Hecht, A., Kamm, R., Weinfurther, K., & Valencia, J. (2023, December). List Mode Inference Using Linear Classifiers for Nuclear Arms Control Verification, ESARDA Bulletin - The International Journal of Nuclear Safeguards and Non-proliferation, 65, 10-18. https://doi.org/10.3011/ESARDA.IJNSNP.2023.2

ESARDA 65 THMB_B_65-2

Files

19 MAY 2023
List Mode Inference Using Linear Classifiers for Nuclear Arms Control Verification
English
(635.36 KB - PDF)
Download