A Review of Anomaly Intrusion Detection Systems in IoT using Deep Learning Techniques

Abstract

The rise of the Internet of things (IoT) provides an intelligent environment. However, it demands a higher security effort due to its vulnerabilities. A number of studies have emphasized the area of anomaly intrusion detection and its use in various types of applications. The development of a robust anomaly intrusion detection system relies heavily on understanding complex structures from noisy data, identifying dynamic anomaly patterns, and detecting anomalies with insufficient labels. Therefore, an advanced approach of deep learning techniques is required for the purpose of achieving improved performance of anomaly detection rather than the unconventional approaches of shallow learning. However, the immense adaptations of IoT in major devices result in an increase in data usage and higher computational requirements. Hence, this work highlights a review on anomaly intrusion detection utilizing deep learning approaches with a focus on resource-constrained devices’ application within the domain of IoT in real-world challenges. Based on findings, the performance of deep learning in anomaly detection is proven to be superior in regard to accuracy detection and false alarm rate. However, further work on deep learning techniques should be carried out to ensure the robustness of IDS.