Divergence-Based Adaptive Aggregation for Byzantine Robust Federated Learning
Summary
This research addresses two problems in federated learning (a training approach where multiple computers work together to improve a shared AI model without sending raw data to one central location): client drift (when different computers' local updates diverge from each other due to unequal data) and Byzantine attacks (when some computers send false or malicious updates to sabotage the model). The paper proposes two frameworks called DRAG and BR-DRAG that use a metric called divergence of degree to measure how far off-track each computer's updates are, then automatically correct them through a process called linear calibration, with BR-DRAG adding a vetted dataset at the main server to defend against malicious actors.
Solution / Mitigation
DRAG addresses client drift by designing a reference direction and divergence metric to quantify deviation of local updates, allowing each worker to align its local update via linear calibration without extra communication cost. BR-DRAG extends this by maintaining a vetted root dataset at the server to produce trusted reference directions that workers' updates are calibrated against, thereby mitigating divergence caused by malicious attacks.
Classification
Related Issues
Original source: http://ieeexplore.ieee.org/document/11579423
First tracked: July 21, 2026 at 08:04 PM
Classified by LLM (prompt v3) · confidence: 85%