Deep Learning Models for Intrusion Detection Using a Text-to-Image Representation Approach

Abstract

Intrusion Detection Systems (IDS) play a critical role in safeguarding modern digital infrastructures, especially in automotive and IoT environments where cyber threats are increasingly prevalent. Traditional IDS methods often struggle with limited adaptability, high false alarm rates, and performance inefficiencies in real-time, resource-constrained settings. To address these challenges, this paper proposes a novel intrusion detection framework that integrates text-to-image conversion of CAN (Controller Area Network) traffic data with deep learning models. The structured textual dataset is transformed into 9×9×3 RGB image representations, enabling the application of advanced CNN-based models such as EfficientNetB0 and InceptionV3. These models are selected for their computational efficiency and high classification accuracy. Experiments conducted on the OCSLab Car-Hacking Dataset demonstrate that the proposed approach significantly outperforms traditional machine learning models in terms of accuracy, F1 score, and computational efficiency. This framework presents a scalable and real-time solution for intelligent and secure vehicular systems.