Levy Flight Strategy with Chicken Swarm Optimization Algorithm for Real-Time Anomaly Detection in Cybersecurity

The cybersecurity field has witnessed several Intrusion Detection System (IDS) and Anomaly Detection, which are essential for identifying malicious activities in network traffic. However, growing volume and diversity of modern network attacks pose challenges, impacting the accuracy of these detection systems. This paper, Levy Flight strategy with Chicken Swarm Optimization Algorithm (LCSO) and Gradient-Boosted Decision Tree (GBDT) techniques achieve better accuracy. GBDT builds decision trees iteratively and helps in reducing bias and variance, leading to high accuracy models, especially with feature selected by using CSO. The CSO algorithm applied to various types of data, making it a flexible feature selection method for different cybersecurity. Initially data obtained from KDD-99 and NSL-KDD dataset and normalization is essential to prevent feature with large range from dominating the learning process. The feature selection using LCSO algorithm focusing on the most relevant features, helps improved efficiency achieve classification accuracy. The GBDT build better classifiers that handle both normal and anomalous pattern efficiently, especially with optimized feature set. The proposed LCSO-GBDT algorithm achieves better accuracy of 99.15% on KDD and 99.12% on NSL-KDD dataset when compared to existing techniques Deep Neural Network (DNN) and Bat Algorithm approach.