Adversarial Forecasting for Command and Control Decisions

• Activity Number: 27 - SDNS Speed Session
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 1:30 PM to 3:20 PM
• Sponsor: Section on Statistics in Defense and National Security
• Author(s): Tahir Ekin* and William N. Caballero and David Rios Insua
• Companies: Texas State University and Air Force Institute of Technology and ICMAT-CSIC
• Keywords: machine learning; adversarial risk analysis; adversarial machine learning; adversarial forecasting; military decision making; command and control decisions

Abstract

All command and control (C2) decisions are founded upon information. Direct manipulation of this information can significantly alter a C2 system’s performance, thereby impacting the quality of the resulting decisions. One such example relates to the targeting of forecasting algorithms that predict operational conditions. Although the forecasting model provides legitimate output, the underlying data may be corrupted by a data-fiddler. To address such challenges, we propose an adversarial hidden Markov Model that maximizes expected performance under adversarial risk analysis assumptions.