ENHANCING SECURITY RESILIENCE: DYNAMIC DRIFT DETECTION IN CLOUD-BASED INTRUSION DETECTION USING THE HYBRID MODEL

Security and trust are seriously challenged by intrusion in cloud computing. The unpredictability of data patterns is a significant barrier that makes it challenging for current machine learning (ML) models to reliably identify and categorize intrusion attacks over an extended period of time. To assess the efficacy of a hybrid model for real-time anomaly detection and adaptive learning, this study combines the XGBoost classifier with the Adaptive Windowing Detection of Drift (AWDD) model. The goal of the hybrid model is to locate drift in streams of continuous data. We evaluate its performance by means of an extensive analysis including diverse metrics and visualizations. According to our research, the XGBoost and AWDD model working together minimizes false negatives while accurately identifying positive cases. The model exhibits strong recall and precision, with an F1-score of 99.6 percent and a precision rate of 99.3 percent. Despite a few false positives, the model continues to function well. The model’s remarkable AUC (area under the curve) value of 0.94 on the ROC curve indicates how accurate it is at making predictions. Additionally, the accuracy trend plot shows a steady rising trend, indicating the adaptability of the hybrid model to shifting data distributions. These findings highlight the hybrid model’s value as a reliable resource for anomaly identification in real time across a range of industries and applications.