AN EFFECTIVE APPROACH FOR INTRUSION DETECTION IN IOT ENVIRONMENT

Abstract

The Internet of Things enables connectivity among heterogeneous physical devices and supports real-time applications in transportation, healthcare, military systems, and agriculture. However, extensive use of communication protocols exposes IoT networks to serious security threats, while traditional signature-based and rule-based intrusion detection techniques remain ineffective against evolving and unknown attacks. Accurate analysis of network traffic behavior is therefore essential for secure IoT operations. This paper presents an Intrusion Detection System (IDS) that is based on a deep learning framework that integrates feature selection based on correlation with a convolutional neural network (CNN) for anomaly detection in a network. Initially, the most informative traffic features are selected using a linear correlation-based extraction strategy to reduce redundancy and improve learning efficiency. The refined feature set is then processed by a convolutional neural network to automatically learn discriminative patterns associated with malicious activities. The proposed system supports both binary classification and multiclass classification for anomaly detection and for identifying different categories of attacks. The proposed approach is evaluated using multiple benchmark datasets representing diverse network environments. For comparison, several conventional machine learning classifiers, including linear discriminant analysis, k-nearest neighbors, classification and regression trees, and support vector machines, are trained using the same feature extraction strategy. Experimental evaluation confirms a significant improvement in the proposed model, which improves security performance in IoT environments.