AI Sec Watch: A Security Intelligence Platform for AI Systems

Luu, T.J.

FedDOT: Defending Federated Learning Against Overwhelming Targeted Attacks

inforesearchPeer-Reviewed

securityresearch

Source: IEEE Xplore (Security & AI Journals)March 23, 2026

Summary

Federated learning (FL, a system where multiple participants train AI models on their own data and share only the updates rather than raw data) is vulnerable to targeted attacks where malicious participants send bad updates to corrupt the shared model. This paper proposes FedDOT, a defense framework that uses two graph-based detection algorithms (maximum spanning tree-based attacker detection and densest graph-based attacker detection) to identify and filter out malicious updates, even when attackers make up more than half of the participants.

Solution / Mitigation

The FedDOT framework implements two defense algorithms: maximum spanning tree-based attacker detection (MST-AD) and densest graph-based attacker detection (density-AD), which leverage correlation between weight updates and graph theory concepts to identify and remove malicious updates from the aggregation process. According to the experimental results presented, FedDOT achieves an attack success rate of less than 10% against targeted attacks while maintaining an accuracy drop of less than 2%.