Enhancing Web Server Security Using Deepfocusguard Based Skydive Optimization For AI-Based Ddos Prediction And Prevention

Modern web servers face an increasing threat from Distributed Denial-of-Service (DDoS) attacks, which can overwhelm resources, disrupt services, and compromise availability, yet traditional defense mechanisms often fail to cope with high traffic volume, evolving attack patterns, and adaptive adversaries. To overcome these challenges, this research proposes DeepFocusGuard Based SkyDive Optimization and is developed for accurate DDoS prediction, efficient prevention, and enhanced web server security. Initially the dataset transforms through data preprocessing, including feature transformation to integer format, handling of missing values to ensure data completeness, and label encoding to convert categorical attributes into machine-readable form. Feature selection is performed using correlation filtering to eliminate redundant attributes, followed by Adaptive Synthetic Sampling Approach (ADASYN) oversampling to balance traffic classes between normal and malicious flows. The processed data is then classified using the DeepFocusGuard framework, where Convolutional Neural Networks (CNN) are employed to extract spatial features from network traffic, Bidirectional Long Short-Term Memory (BiLSTM) captures temporal dependencies across sequential traffic patterns, and the Attention mechanism highlights critical traffic behaviors, collectively enhancing DDoS detection accuracy. To further strengthen performance, SkyDive Optimization developed by combining Pelican Optimization (PO) and Aquilo Optimizer (AO) is applied for adaptive hyperparameter tuning, improving convergence, reducing overfitting, and ensuring generalization across diverse traffic conditions. Experimental results demonstrate that the proposed model achieves superior accuracy, precision, recall, F1-score, AUC-ROC, AUC-PR, Attack Mitigation, False Blocking, and Server Availability compared to baseline traditional defenses and state-of-the-art methods, thereby ensuring stronger DDoS prevention and improved web server security. Future research will extend this work by incorporating explainable AI modules for interpretability, real-time adaptive learning against novel attacks, and large-scale deployment in heterogeneous web server environments.