An improved Network Intrusion Detection System using a Novel Hybrid Genetic Coati-Pelican Optimization Algorithms with an Ensemble Fuzzy c-means and CNN-RF

Network intrusion remains a critical security concern in the modern cyber world, necessitating innovative approaches to enhance network confidentiality. In this paper, we introduce a novel Hybrid Genetic Coati-Pelican Optimization Algorithm (hGCPOA) designed to improve feature selection by expanding population size, increasing generation count, and enhancing individual exploration characteristics. Leveraging fitness functions, Holt’s approach, and Gaussian mutation, our methodology incorporates a customized fitness function to refine the search for optimal solutions. This step ensures that the hGCPOA converges towards the most significant features while maintaining genetic diversity, thereby avoiding premature convergence. The optimally selected features undergo analysis using an advanced anomaly detection framework comprising an ensemble model of Fuzzy c-means clustering, Densely Convolutional Neural Networks (DCNN), and Random Forest (RF) algorithms. Fuzzy c-means clustering partitions the data into categories indicative of normal and anomalous traffic patterns, while DCNN executes multi-class parameter analysis, leveraging its robust feature representation capabilities. RF enhances this ensemble by offering strong classification performance and increasing predictive accuracy, effectively differentiating between benign and malicious activities. Evaluation of our model on NSL-KDD and CIC-IDS2017 datasets demonstrates exceptional performance across accuracy, precision, recall, and F1 score metrics, achieving a classification accuracy of 99.99%. Our approach underscores the significance of hybrid optimization in developing resilient and adaptive Network Intrusion Detection Systems (NIDS) to combat evolving cyber threats effectively.