Noah Oikarinen
Department of Electrical and Computer Engineering, University of Massachusetts, Dartmouth, MA, USA.

Liudong Xing
Department of Electrical and Computer Engineering, University of Massachusetts, Dartmouth, MA, USA.

DOI https://doi.org/10.33889/IJMEMS.2025.10.6.078

  ;

Abstract

The voting-based redundancy mechanism is widely used as a fault-tolerant technique across various engineering and computing domains. However, this technique is vulnerable to colluding attacks where multiple malicious resources can collaborate to produce identical wrong results to potentially fail a task execution, posing significant threats to system integrity, performance, and reliability. This research introduces a heterogeneous multi-agent colluding attack defense system that employs a two-stage spot checking strategy within a credibility-based framework designed to detect and mitigate the impact of coordinated adversarial behavior within diverse agent environments. The proposed framework employs a dual-role architecture, consisting primarily of spotter and resource agents. Spotters are responsible for monitoring and evaluating the credibility of resource agents based on their performance and voting patterns. Resource agents that fail the first spot-check cannot participate in the voting process, while a second spot check targets a randomly selected agent from the majority vote, revoking the credibility of any colluding participants. The system dynamically and optimally allocates these agents under users’ predefined cost constraints, balancing resource utilization and defense efficiency. By strategically adjusting the credibility scores of resource agents, the proposed defense mechanism adapts to ensure sustained system performance and reliability. Experimental studies are conducted to demonstrate effectiveness of the proposed heterogeneous multi-agent system in defending against colluding attacks in voting-based computing environments. The impacts of several key model parameters on system performance are also investigated.

Keywords- Colluding attack, Multi-agent system, Reliability, Spot-checking, Voting.

Citation

Oikarinen, N., & Xing, L. (2025). A Heterogeneous Multi-Agent Colluding Attack Defense System. International Journal of Mathematical, Engineering and Management Sciences, 10(6), 1640-1657. https://doi.org/10.33889/IJMEMS.2025.10.6.078.