Home▸Machine Learning & Neural Networks▸Secure Aggregation in Federated Learning: Pairwise Masks and Dropout

🔐 Secure Aggregation in Federated Learning: Pairwise Masks and Dropout

Federated clients add pairwise random masks that cancel in the server sum, so the server learns only the total of their private updates. See how dropouts leave masks uncancelled, how Shamir shares let the survivors repair the sum, the exact modular arithmetic, and the odds that a round succeeds. A toy model, not real cryptography.

Machine Learning & Neural Networks3DAdvanced60 FPS📱 Mobile-adapted🟠 Blender⇄ 2D version
Secure Aggregation in Federated Learning: Pairwise Masks and Dropout ↗ Open standalone
⚙ Under the hood

Federated clients add pairwise random masks that cancel in the server sum, so the server learns only the total of their private updates. See how dropouts leave masks uncancelled, how Shamir shares let the survivors repair the sum, the exact modular arithmetic, and the odds that a round succeeds. A toy model, not real cryptography.

secure aggregationfederated learningpairwise maskingShamir secret sharingdropoutmodular arithmeticprivacymachine learning

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install

What did you find?

Add reproduction steps (optional)