Enhanced Anomaly Detection in Automotive Systems Using SAAD: Statistical Aggregated Anomaly Detection

Dacian Goina,Eduard Hogea,George Maties

2024-06-11

Abstract:This paper presents a novel anomaly detection methodology termed Statistical Aggregated Anomaly Detection (SAAD). The SAAD approach integrates advanced statistical techniques with machine learning, and its efficacy is demonstrated through validation on real sensor data from a Hardware-in-the-Loop (HIL) environment within the automotive domain. The key innovation of SAAD lies in its ability to significantly enhance the accuracy and robustness of anomaly detection when combined with Fully Connected Networks (FCNs) augmented by dropout layers. Comprehensive experimental evaluations indicate that the standalone statistical method achieves an accuracy of 72.1%, whereas the deep learning model alone attains an accuracy of 71.5%. In contrast, the aggregated method achieves a superior accuracy of 88.3% and an F1 score of 0.921, thereby outperforming the individual models. These results underscore the effectiveness of SAAD, demonstrating its potential for broad application in various domains, including automotive systems.

Machine Learning

What problem does this paper attempt to address?

The paper proposes a new method called Statistical Aggregated Anomaly Detection (SAAD) to address the anomaly detection problem in automotive systems. In the automotive industry, anomaly detection is crucial for ensuring the reliability and safety of systems, especially in autonomous vehicles. With the development of machine learning techniques, particularly the application of deep learning models, the accuracy and efficiency of anomaly detection have been significantly improved.

The SAAD method combines advanced statistical techniques and machine learning by incorporating dropout layers into a fully connected network (FCN) to enhance its performance. The effectiveness of SAAD was validated using real sensor data from hardware-in-the-loop (HIL) testing environments in the automotive industry. The experiment results showed that the standalone statistical method achieved an accuracy of 72.1%, the standalone deep learning model achieved an accuracy of 71.5%, while the combined SAAD method achieved an accuracy of 88.3% and an F1 score of 0.921, outperforming the standalone models.

The advantage of SAAD lies in improving the accuracy and robustness of FCN in real-time applications. It can better identify anomalies in automotive systems by integrating statistical methods and deep learning, enabling real-time monitoring of sensor data streams to proactively detect potential faults or network security attacks.

The paper reviews related work, including intrusion detection systems, statistical learning methods, and the application of deep learning models in different domains, and emphasizes the potential of combining statistical methods with machine learning. SAAD achieves more efficient anomaly detection through a methodology that involves manual annotation, deep learning model development, and result aggregation. The experiment data comes from HIL testing environments, demonstrating the advantages of SAAD in improving detection accuracy and robustness.