MineDetect combines history-averaged gradient similarity, variance, and distance thresholds to flag sign-flip, noise, and unreliable clients in federated learning, but its headline accuracy gain is not borne out by the reported tables.
Privacy and robustness in federated learning: Attacks and defenses,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Detecting Untargeted Attacks and Mitigating Unreliable Updates in Federated Learning for Underground Mining Operations
MineDetect combines history-averaged gradient similarity, variance, and distance thresholds to flag sign-flip, noise, and unreliable clients in federated learning, but its headline accuracy gain is not borne out by the reported tables.