Federated learning + iDLG gradient inversion attack + Central DP defense + Gradio demo. The honest finding: naive Central DP collapses utility on small federations (Gaussian-mechanism curse of dimensionality). Production fixes (DP-SGD, Opacus) documented.
security pytorch gradio differential-privacy federated-learning privacy-preserving-ml gradient-inversion dlg-attack
-
Updated
Apr 26, 2026 - Python