Back to the MIT repository
7. AI System Safety, Failures, & Limitations2 - Post-deployment

Swarm Attacks

Swarm Attacks. The need for multi-agent security is foreshadowed by attacks today that benefit from the use of many decentralised agents, such as distributed denial-of-service attacks (Cisco, 2023; Yoachimik & Pacheco, 2024). Such attacks exploit the massive collective resources of individual low- resourced actors, chained into an attack that breaks the assumptions of bandwidth constraints on a single well-resourced agent.

Source: MIT AI Risk Repositorymit1243

ENTITY

1 - Human

INTENT

1 - Intentional

TIMING

2 - Post-deployment

Risk ID

mit1243

Domain lineage

7. AI System Safety, Failures, & Limitations

375 mapped risks

7.6 > Multi-agent risks

Mitigation strategy

1. Implement Online Ensemble Learning and User/Entity Behavior Analytics (UEBA) within a decentralized architecture to enable real-time, adaptive classification of coordinated anomalous behavior and concept drift, thereby pre-empting swarm emergence. 2. Enforce Decentralized, Resilient Communication Protocols with Robust Cryptographic Authentication to ensure agent identity integrity, secure inter-agent communication channels, and prevent spoofing or protocol manipulation. 3. Employ Dynamic Load Shifting and Adaptive Clustering/Assignment mechanisms, coupled with Architectural Redundancy (e.g., multi-path routing) to absorb or re-allocate computational and communication load away from targeted components.