Abstract:Multi-agent systems rely on communication for information sharing and action coordination, which exposes a vulnerability to attacks. We investigate single-victim communication perturbation attacks against Multi-Agent Reinforcement Learning-trained systems and propose methods that use gradient information from the Jacobian to identify which messages, agent, and timesteps are most susceptible to attack and have the greatest impact on the system. We enhance these methods with two proposed adversarial loss functions that trade-off attack success for attack impact which also create more effective perturbations. We empirically demonstrate the effectiveness of our methods against two different multi-agent communication methods in navigation, PredatorPrey, and TrafficJunction environments. Our results show that our novel message selection method achieves a similar or greater impact than random message selection across almost all tested scenarios. Our victim selection, message selection, tempo, and loss functions improve attack effectiveness in half of the thirty scenarios we tested.
| Comments: | Full version of the Extended Abstract presented at AAMAS 2026 |
| Subjects: | Machine Learning (cs.LG); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2605.13170 [cs.LG] |
| (or arXiv:2605.13170v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.13170 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Maxwell Standen [view email]
[v1]
Wed, 13 May 2026 08:32:30 UTC (43 KB)