FedSecureFormer: A Fast, Federated and Secure Transformer Framework for Lightweight Intrusion Detection in Connected and Autonomous Vehicles

This addresses security threats for connected and autonomous vehicles, but appears incremental as it combines existing methods (transformers and federated learning) for a specific application.

The paper tackles intrusion detection in Connected and Autonomous Vehicles by developing a lightweight encoder-only transformer framework using Federated Learning, achieving unspecified performance metrics.