Machine-learning-based onboard detection of SPaT-specific attacks

Develop machine-learning-based intrusion-detection methods for identifying SPaT-specific attacks from the onboard perspective of connected vehicles.

Background

The paper distinguishes onboard connected-vehicle detection of SPaT attacks from infrastructure-side traffic-signal intrusion detection and from V2V Basic Safety Message misbehavior detection. It identifies a gap in methods that use machine learning to detect malicious SPaT messages at the receiving vehicle.

The SPADE contribution addresses the dataset gap by providing labelled, multi-modal simulated data containing SPaT, camera-perception, and V2V information. However, the development and comparative evaluation of machine-learning intrusion-detection models using this data remain unresolved, as the paper explicitly characterizes ML-based onboard detection of SPaT-specific attacks as still open.

References

Abdel Hakeem and Kim survey ML, federated learning, and edge AI for V2X IDS broadly, confirming that ML-based onboard detection of SPaT-specific attacks remains open and that no existing dataset addresses it.

— SPADE: SPaT Attack Detection from the Connected Vehicle's Perspective  (2609.02741 - Novo et al., 2 Sep 2026) in Section 2, subsection “Traffic Signal Security”