Home▸Cybersecurity▸Byzantine-Robust Federated Learning Defense (2D)

Byzantine-Robust Federated Learning Defense (2D Vector-Field View)

Interactive 2D federated-learning security simulator: watch malicious clients send poisoned gradient updates toward a shared global model, then compare a naive FedAvg mean against a Byzantine-robust coordinate-wise median defense that filters the attack out, with a live vector field plus a round-by-round distance/impact chart.

Cybersecurity2DAdvanced60 FPS📱 Mobile-adapted⇄ 3D version

Details

CategoryCybersecurity

RenderingCanvas 2D

DifficultyAdvanced

Target FPS60

ControlsSliders set the number of malicious clients (0-3 of 6) and their attack strength, buttons switch the server's aggregation rule between naive mean and Byzantine-robust coordinate-wise median and reset training, and a speed slider controls rounds per second; the vector-field panel supports drag-to-pan and scroll/pinch-to-zoom.

federated learningML securityByzantine robustnessmodel poisoninggradient aggregationAI defense

Related simulations

FedAvg Loss-Curve Drift: 2D Federated Model AggregationInteractive 2D federated-learning simulator: five clients report scalar weight updates onto…

Federated Learning SimulationThis simulation demonstrates the concept of federated learning, where machine learning…

Gradient Leakage Attack on Federated LearningWatch an attacker reconstruct a client's private 8x8 image from a single leaked model… ⊞ All 534 in Cybersecurity →

Federated learning trains one shared model from many clients' local gradient updates without ever centralizing their raw data — but that also means the server must trust updates it cannot inspect. This 2D simulator renders six clients arranged around a central global model: each round every client sends a local update vector, one or more can be set malicious and send a poisoned update scaled away from the shared true optimum, and the server aggregates all six into the next model position. A draggable, zoomable vector-field view shows every gradient arrow live, while a companion strip chart tracks the model's distance from the true optimum and the measured poisoning impact on the aggregate round by round. Switch the aggregation rule between a naive mean — which a single unbounded malicious update can drag anywhere — and a Byzantine-robust coordinate-wise median, which outvotes a minority of poisoned updates instead of averaging them in.

⚙ Under the hood

Interactive 2D federated-learning security simulator: watch malicious clients send poisoned gradient updates toward a shared global model on a draggable, zoomable vector field, then compare a naive FedAvg mean against a Byzantine-robust coordinate-wise median defense that filters the attack out, with a live distance/impact strip chart.

federated learningML securityByzantine robustnessmodel poisoninggradient aggregationAI defense

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install