Threat Assessment Model for Heterogeneous Multi-Agent Systems (14291-16)

Description

Heterogeneous Multi-Agent Systems (MAS) that span systems-of-systems of multi-domain robot fleets operating over open networks in dynamic and adversarial environments, are exposed to complex and evolving attack surfaces. Existing studies in the field of robotics security largely focus on single-domain Multi-Robot Systems (MRS) such as swarms of unmanned aerial vehicles, and threat model frameworks like STRIDE (Spoofing Tampering Repudiation Information disclosure Denial of service Elevation of Privilege) and PASTA (Process for Attack Simulation and Threat Analysis) are popularly used to mitigate potential threats. However, there is still a lack of study of a systematic threat model and assessment tailored to heterogeneous MAS. This gap is critical since the heterogeneous MAS have different capabilities given their different communications, sensors, actuators, payloads, hardware, and software. To this end, this paper proposes a threat model framework for heterogeneous multi-agent systems that explicitly targets this gap. Accordingly, this research (i) defines a domain-independent four-layer reference architecture, spanning the Agent, Coordination, Communication, and Human layers, that captures the assets, interfaces, and trust boundaries of a heterogeneous MAS; (ii) develops a structured assessment methodology that combines the affected layer, the target criticality, the event severity, the effect impact, and the propagation scope into a single quantitative threat assessment score; and (iii) applies the model to a coastal mine detection scenario with a heterogeneous fleet of UAVs, UGVs, and USVs, showing how threats across the four layers and the mission phases are identified and then ranked by that score. The proposed framework for heterogeneous MAS can serve as a security design guidance beyond single-platform case studies.

Description complémentaire

Conference 14291: Autonomous Systems for Security and Defence III

Période	15 sept. 2026
Titre de l'événement	SPIE Sensors + Imaging: SPIE Environmental Remote Sensing and SPIE Security + Defence
Type d'événement	Une conférence
Emplacement	Edinburgh, Royaume-UniAfficher sur la carte
Degré de reconnaissance	International