Deep Learning-based Embedded Intrusion Detection System for Automotive CAN

Shashwat Khandelwal,Eashan Wadhwa,Shreejith Shanker

DOI: https://doi.org/10.1109/ASAP54787.2022.00023

2024-01-19

Abstract:Rising complexity of in-vehicle electronics is enabling new capabilities like autonomous driving and active safety. However, rising automation also increases risk of security threats which is compounded by lack of in-built security measures in legacy networks like CAN, allowing attackers to observe, tamper and modify information shared over such broadcast networks. Various intrusion detection approaches have been proposed to detect and tackle such threats, with machine learning models proving highly effective. However, deploying machine learning models will require high processing power through high-end processors or GPUs to perform them close to line rate. In this paper, we propose a hybrid FPGA-based ECU approach that can transparently integrate IDS functionality through a dedicated off-the-shelf hardware accelerator that implements a deep-CNN intrusion detection model. Our results show that the proposed approach provides an average accuracy of over 99% across multiple attack datasets with 0.64% false detection rates while consuming 94% less energy and achieving 51.8% reduction in per-message processing latency when compared to IDS implementations on GPUs.

Cryptography and Security,Machine Learning

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to implement an efficient, low - power - consumption and real - time Intrusion Detection System (IDS) in the in - vehicle network (such as CAN bus). With the increasing complexity of the in - vehicle electronic systems and the development of autonomous driving and active safety functions, the security threats faced by these systems are also increasing. In particular, the traditional CAN network lacks built - in security measures, enabling attackers to observe, tamper with and modify the information shared on these broadcast networks. Although various intrusion detection methods have been proposed to detect and deal with these threats, and machine - learning models have proven to be very effective, the deployment of these models requires high - performance processors or GPUs, which not only increases the cost, but also increases the energy consumption and processing latency.

To solve these problems, the paper proposes a hybrid ECU (Electronic Control Unit) method based on FPGA, which transparently integrates IDS functions through dedicated off - the - shelf hardware accelerators. The accelerator implements a Deep - Convolutional Neural Network (Deep - CNN) intrusion detection model. This method aims to provide high - precision intrusion detection while significantly reducing energy consumption and processing latency, making it suitable for real - time applications in the in - vehicle environment. Specifically, the goals of the paper include:

1. **Improve detection accuracy**: The model proposed in the paper achieves an average accuracy of over 99% on multiple attack data sets, with a false detection rate of 0.64%.
2. **Reduce energy consumption**: Compared with the GPU implementation, the proposed scheme reduces energy consumption by 94%.
3. **Reduce processing latency**: The processing latency of each message is reduced by 51.8%.
4. **Achieve real - time detection**: Through the tightly - coupled FPGA accelerator, real - time detection of CAN network messages is achieved to ensure vehicle safety.

In conclusion, this paper aims to enhance the security of in - vehicle networks through an innovative hardware acceleration method while maintaining low power consumption and low latency to adapt to the increasingly complex electronic systems of modern automobiles.