AN IMPROVED DEEP LEARNING BASED INTRUSION DETECTION SYSTEM FOR IoT NETWORKS

Abstract

The Internet of Things (IoT) has changed the way people live by connecting smart devices smoothly in many areas like healthcare, smart homes, and industrial automation. But the fast growth of the Internet of Things has caused a big rise in the number of connected devices and the amount of data being created. By 2025, it is expected that about 41.6 billion devices will generate around 79.4 zettabytes of data. Deep learning-based intrusion detection systems have shown great potential for understanding complicated traffic pattern. However, Most existing models often suffer from overfitting, poor generalization, and high computational cost, making them unsuitable for resource-constrained IoT environments. This study proposes an improved Intrusion Detection System that integrates a Recurrent Neural Network (RNN) with ensemble classifier combining Support Vector Machine (SVM) and Logistic Regression (LR). The RNN captures temporal dependencies within network traffic, while the ensemble classifier enhances generalization and reduces overfitting. The model is evaluated using the CICIDS2017 dataset. Experimental results demonstrate an accuracy of 88%, with reduced computational complexity compared to traditional methods. The proposed approach offers a practical solution for real-time IoT security applications, particularly in resource-constrained environments.