Deception Against Game-Theoretic Learning in Complex Adaptive Systems

Vamvoudakis, Kyriakos G.

Permalink

https://hdl.handle.net/2142/130263 Copy

Description

Title

Deception Against Game-Theoretic Learning in Complex Adaptive Systems

Author(s)

Vamvoudakis, Kyriakos G.

Issue Date

2025-09-17

Keyword(s)

• Game theory
• Security
• Cyber-physical systems

Abstract

Complex adaptive systems (CAS) are characterized by their intricate complexity and diverse components, both digital and analog, which communicate through multiple channels. These systems are prevalent in various applications, making them attractive targets for adversarial attacks aimed at causing damage, confusion, and deception. Deception acts as a strategic defense mechanism against adversaries lacking crucial information, forcing them to adopt less effective strategies to the defender’s benefit. When an adversary seeks to determine the optimal attack on an unfamiliar system, a defender with thorough knowledge of the system’s dynamics can introduce deceptive inputs to mislead the adversary. This work aims to design these deceptive inputs to guide the adversary towards learning a pre-selected, suboptimal attack. Specifically, we demonstrate how to corrupt the dynamics of a learning algorithm to mislead it into converging on a suboptimal policy, and how to directly attack the CAS to inflict maximum damage while remaining undetected. This challenge involves solving a coupled algebraic Riccati and Lyapunov equation, which are analytically complex. However, a block successive overrelaxation algorithm is employed to obtain their numerical solution, with proven convergence under specific conditions. Simulations on a benchmark aircraft showcase the algorithm’s effectiveness in steering adversaries towards less damaging attacks.

Publisher

Allerton Conference on Communication, Control, and Computing

Series/Report Name or Number

2025 61st Allerton Conference on Communication, Control, and Computing Proceedings

ISSN

2836-4503

Type of Resource

Text

Genre of Resource

Conference Paper/Presentation

Language

eng

Handle URL

https://hdl.handle.net/2142/130263

Copyright and License Information

Copyright 2025 is held by Kyriakos G. Vamvoudakis.

Owning Collections