MS-CFFS: Multistage Coarse and Fine Feature Selecton for Advanced Anomaly Detection in IoT Security Networks
Mohammed Sayeeduddin Habeeb, Tummala Ranga Babu · International Journal of Electrical and Electronics Research · 2024
In recent years, the concept of Internet-of-Things (IoT) has increased in popularity, leading to a massive increase in both the number of connected devices and the volume of data they handle. With IoT devices constantly collecting and sharing large quantities of sensitive data, securing this data is of major concern, especially with the increase in network anomalies. A network-based anomaly detection system serves as a crucial safeguard for IoT networks, aiming to identify irregularities in the network entry point by continuously monitoring traffic. However, the research community has contributed more to this field, the security system still faces several challenges with detecting these anomalies, often resulting in a high rate of false alarms and missed detections when it comes to classifying network traffic and computational complexity. Seeing this, we propose a novel method to increase the capabilities of Anomaly Detection in IoT. This study introduces the deep learning (DL) based Multistage Coarse and Fine Feature Selection (MS-CFFS), to improve anomaly detection techniques devised for IoT security frameworks. The proposed feature section is done in two stages. The MS-CFFS, utilizing a deep learning-based dual-stage feature selection, substantially improves NIDS efficacy. The results confirm MS-CFFS's outstanding classification accuracy at 99.93%, with a remarkably low FAR of 0.05% and FNR of 0.11%. These achievements stem from refining the feature set to 28 pivotal features, thus notably cutting computational complexity without sacrificing precision. Furthermore, a comparative analysis with leading-edge approaches validates the preeminence of our proposed MS-CFFS in the domain of network security.