Enhancing Autonomous Vehicle Safety through LSTM-AE Based Anomaly Detection

Abstract

Vehicles are essential in our lives, providing popular private transportation. Despite their comfort, they pose road safety risks. Autonomous vehicles, a technological advancement, mitigate these risks with sensors and algorithms, reducing human error. However, they introduce cybersecurity challenges, necessitating urgent research to detect and mitigate potential threats to ensure safety. This work suggested a novel anomaly detection technique, the Long Short-Term Memory based Autoencoder (LSTM - AE), to address the issue. First, use an autoencoder to extract features from the autonomous vehicle data and compress them into a latent representation in order to train the model. The LSTM network receives the compressed features and uses them to identify any linked relationships between the features. The output is a reconstruction of the features. Anomalies depending on the output's reconstruction loss are found using an anomaly score. To evaluate effectiveness of the proposed framework, various evaluation metrics such as accuracy, precision, recall, F1 score and ROC curve analysis are employed. The experimental findings are compared with existing models, such as Convolution Neural Networks (CNN), Recurrent Neural Networks (RNN) and Deep Reinforcement Learning (DRL), and Deep Belief Networks (DBN). With an accuracy of 92% and a larger area under the receiver operating characteristic curve than alternative approaches, the comparative findings show that the suggested model outperforms the others in anomaly detection, making it a useful tool for autonomous vehicle anomaly detection. Finally, it is determined that the model's output can be used to intelligently identify and prevent cyberattacks against autonomous cars.