Enhancing IoT Network Security: A Double Decker Convolutional Neural Network with Brown-Bear Optimization for Intrusion Detection

The extensive adoption of Internet of Things (IoT) networks has exponentially increased cybersecurity threats, necessitating effective intrusion detection mechanisms to secure connected devices. Traditional security mechanisms fail to detect sophisticated cyberattacks as IoT environments are dynamic in nature. Existing intrusion detection techniques suffer from high false alarms and low computational efficiency, making them impractical to apply in resource-constrained IoT networks. To overcome these challenges, this paper presents Enhancing IoT Network Security: A Double Decker Convolutional Neural Network with Brown-Bear Optimization for Intrusion Detection (BrBO-DD-CNN), which yields high detection efficiency and accuracy. This paper utilizes data from the NSL-KDD and BoT-IoT datasets, which are IoT network traffic and attack types. Raw data is preprocessed by Min-Max Zscore Normalization to remove duplicates and complete missing values. Planet Optimization selects significant features, reducing data complexity without losing key information. Preprocessed data is input to a Double Decker Convolutional Neural Network that detects low-level and high-level features to differentiate normal and malicious traffic. Brown-Bear Optimization optimizes model parameters, enhancing convergence and detection performance. This research yields an accuracy of 99.98% for NSL-KDD and 99.95% for BoT-IoT. Detection rate is 99.97% and 99.93%, and F1-score is 99.96% and 99.92%, respectively, proving its effectiveness. The results guarantee that BrBO-DD-CNN performs better than existing intrusion detection techniques, offering a highly accurate, efficient, and scalable intrusion detection mechanism to secure IoT networks against emerging cyberattacks.