IoT NETWORK SECURITY BASED ON INTRUSION DETECTION SYSTEM USING HYBRID LEARNING ALGORITHMS

Abstract

Recently, the adoption of IoT has increased as it becomes more integrated into our daily lives. Conventional Intrusion Detection Systems (IDS) are struggling to keep up with the Internet of Things' ever-changing security threats. This paper introduces a novel approach to addressing challenges with processing capabilities, storage, and false alarm rates in traditional IDS. Our solution utilizes a CNN to extract features, deriving anomaly detection attributes from IoT device raw data. These features are fed into XGBoost, a model that identifies complex data correlations, to improve intrusion detection. Deep learning and ensemble techniques can protect IoT environments from various attacks. Our method, using the Bot-IOT dataset which includes four types of IoT network attacks, achieves an accuracy of 99.36%. This study strengthens IoT security and resilience against cyberattacks.