Explainable Cybersecurity Intrusion Detection Framework using Machine Learning and Ensemble Models
Abstract
Digital communication systems are rapidly growing, and this rapid growth has also brought a lot of risk of attacks and unauthorized network intrusion, making the need for intelligent intrusion detection extremely high. Efficiently detecting complex and evolving attack patterns is often a challenge with traditional approaches to security. To tackle this challenge, the current research proposes a machine learning and explainable artificial intelligence based malicious network activity detection is proposed. A variety of machine learning and ensemble learning models were applied and compared with the metrics of Accuracy, Precision, Recall, F1 score and ROC-AUC. Experimental analysis showed that the ensemble learning techniques achieved good performance in classification in comparison with traditional machine learning models. The Gradient Boosting classifier performed best with an accuracy of 89.72%, a precision of 100%, a recall of 77.36%, an F1-score of 87.23% and an ROC AUC value of 0.8914, which is the highest results among all approaches implemented, showing a good performance in identifying malicious intrusion activities. In addition, explainable artificial intelligence (XAI) techniques like LIME, SHAP, PDP and ICE were employed to enhance the transparency and interpretability of the models by analyzing the most significant cybersecurity attributes driving prediction decisions. The proposed framework is reliable, interpretable and high-performance intrusion detection for modern cybersecurity applications.