Home▸Machine Learning & Neural Networks▸Secure Aggregation in Federated Learning

Secure Aggregation in Federated Learning

Interactive 3D simulator of the secure-aggregation protocol used in federated learning: watch pairwise random masks hide every client's raw update from the server while their sum still cancels out to the exact true aggregate.

Machine Learning & Neural Networks3DAdvanced60 FPS📱 Mobile-adapted⇄ 2D version

Details

CategoryMachine Learning & Neural Networks

RenderingThree.js · WebGL

DifficultyAdvanced

Target FPS60

ControlsSliders set the number of clients and the pairwise mask magnitude, a Run button executes a masked-aggregation round with animated update flights to the server, a dropout button simulates a client vanishing mid-round, and Server-view/Debug-view buttons toggle between the masked values the server actually sees and the raw per-client updates underneath.

federated-learningsecure-aggregationprivacycryptographymachine-learningdistributed-systems

Related simulations

Federated Learning Communication Efficiency Simulator (2D)Watch 10 federated-learning clients compress their model updates with top-K sparsification…

Federated Learning Communication Efficiency SimulatorWatch 10 federated-learning clients compress their model updates with top-K sparsification…

Byzantine-Robust Federated AggregationInteractive 3D simulator of Byzantine-robust federated learning: malicious clients send… ⊞ All 226 in Machine Learning & Neural Networks →

Federated learning trains a shared model across many clients' private data without ever centralising it — but the model updates themselves can still leak information unless the server is prevented from seeing any single client's contribution. This simulator renders each client as a node in 3D space holding a private update vector, applies the real pairwise-masking scheme used in production secure-aggregation protocols, and animates the masked values flowing to a central server. Live readouts compare the true aggregate against what the server reconstructs after the masks cancel, a toggle exposes both the "server view" (masked, individually unreadable) and a "debug view" (raw values), and a dropout button demonstrates why real deployments need secret-sharing to stay correct when a client disappears mid-round.

⚙ Under the hood

Interactive 3D simulator of the pairwise-masking secure-aggregation protocol used in federated learning: client updates are individually hidden from the server, yet their sum still cancels the masks to reveal the exact true aggregate.

federated-learningsecure-aggregationprivacycryptographymachine-learningdistributed-systems

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