Temporal transformer CAN encoder with federated lightweight heads for anomaly detection

Name: Temporal transformer CAN encoder with federated lightweight heads for anomaly detection
Uploaded: 2026-10-11T17:16:26.510960+00:00
Duration: 287 s
Description: The framework detects anomalies in vehicle CAN messages by combining a Temporal Transformer encoder with Federated Lightweight Heads. It processes time series data into fixed-length frames, uses a Transformer to create rich temporal embeddings, and then employs XGBoost for prediction. Federated Learning allows collaborative training across different vehicle ECUs while keeping raw data private.

The framework detects anomalies in vehicle CAN messages by combining a Temporal Transformer encoder with Federated Lightweight Heads. It processes time series data into fixed-length frames, uses a Transformer to create rich temporal embeddings, and then employs XGBoost for prediction. Federated Learning allows collaborative training across different vehicle ECUs while keeping raw data private.

Open the full episode