BLAST: A Stealthy Backdoor Leverage Attack Against Cooperative Multi-Agent Deep Reinforcement Learning-Based Systems

doi:10.1109/tccn.2026.3706546

BLAST: A Stealthy Backdoor Leverage Attack Against Cooperative Multi-Agent Deep Reinforcement Learning-Based Systems

2026-06-23

0

PRE

AI

DOI：10.1109/tccn.2026.3706546

Original

Original request for help

Save

Abstract

En 中文

Recent studies have shown that cooperative multi-agent deep reinforcement learning (c-MADRL) is under the threat of backdoor attacks. Once a backdoor trigger is observed, it will perform malicious actions, resulting in failures or achieving malicious goals. However, existing backdoor attacks suffer from several issues, e.g., instant trigger patterns lack stealthiness, the backdoor is trained or activated by an additional network, or all agents are backdoored. To this end, in this paper, we propose a novel <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\boldsymbol {B}$ </tex-math></inline-formula>ackdoor <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\boldsymbol {L}$ </tex-math></inline-formula>everage <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\boldsymbol {A}$ </tex-math></inline-formula>ttack againST c-MADRL, BLAST, which attacks the entire multi-agent team by embedding the backdoor only in a single agent. Firstly, we introduce adversary spatiotemporal behavior patterns as the backdoor trigger rather than manually injected fixed visual patterns or instant status and control the period to perform malicious actions. This method can guarantee the stealthiness and practicality of BLAST. Secondly, we hack the original reward function of the backdoor agent via unilateral guidance to inject BLAST, to achieve the leverage attack effect that can pry open the entire multi-agent system via a single backdoor agent. We evaluate our BLAST against 3 classic c-MADRL algorithms (VDN, QMIX, and MAPPO) in 2 popular c-MADRL environments (SMAC and Pursuit), and 3 existing defense mechanisms. The experimental results demonstrate that BLAST can achieve a high attack success rate while maintaining a low clean performance variance rate.

Keywords:

Cooperative multi-agent deep reinforcement learning

backdoor attack

unilateral influence

BLAST: A Stealthy Backdoor Leverage Attack Against Cooperative Multi-Agent Deep Reinforcement Learning-Based Systems

Abstract

Journal

Organization