Federated Lightweight Transformer Improves CAN Bus Intrusion Detection

Global: Federated Lightweight Transformer Improves CAN Bus Intrusion Detection

Researchers Devika S, Pratik Narang, and Tejasvi Alladi announced the development of a new intrusion detection system (IDS) for connected and autonomous vehicles in a paper submitted to arXiv on December 30, 2025. The study, titled FedLiTeCAN, introduces a federated lightweight Transformer model designed to provide fast and robust detection of malicious activity on Controller Area Network (CAN) buses.

Study Overview

According to the arXiv abstract, the authors implemented a lightweight Transformer architecture that operates within a federated learning framework, allowing multiple vehicle nodes to collaboratively train the model without sharing raw data. The approach aims to address the latency and resource constraints typical of on‑board vehicle hardware.

Security Context for CAN Bus

CAN buses are the primary communication backbone in modern vehicles, and their lack of built‑in authentication makes them vulnerable to a range of attacks. As vehicles become increasingly connected, the need for efficient, real‑time IDS solutions has grown, prompting research into AI‑driven detection methods.

Model Architecture

The FedLiTeCAN system leverages a compact Transformer design that reduces parameter count while preserving the ability to capture temporal patterns in CAN traffic. By distributing training across participating vehicles, the federated setup minimizes the exposure of sensitive telemetry data.

Implementation and Evaluation

The paper reports that the model was evaluated on standard CAN‑bus intrusion datasets, achieving detection speeds suitable for real‑time deployment. The submission file, listed as 560 KB, contains the full methodological description and experimental results.

Implications for Vehicle Safety

If adopted, the proposed IDS could enhance the security posture of connected and autonomous vehicles, offering a scalable solution that aligns with industry constraints on computational resources and data privacy.

Next Steps and Availability

The authors indicate that future work will explore broader federated scenarios and integration with vehicle‑level security frameworks. The preprint is publicly accessible through arXiv, and the authors have provided links to associated code repositories.

This report is based on information from arXiv, licensed under Academic Preprint / Open Access. Based on the abstract of the research paper. Full text available via ArXiv.