Design of an Improved Model Using Hybrid GA-CNN and DQN for Real Time Cybersecurity Enhancement

The increasing complexity of cyber-attacks requires advanced security methods, as traditional rule-based systems often fail in dynamic environments and zero-day attacks. To address this, we propose a deep learning-based cybersecurity framework for real-time threat detection and mitigation. Using a Hybrid GA-optimized CNN, our model achieves over 95% accuracy in classifying network traffic as benign or malicious. The GA further optimizes the CNN, improving accuracy by 5-10%. A Deep Q-Network is embedded for real-time anomaly detection and self-healing, reducing response times by 30-40% and achieving an 85-90% success rate in threat neutralization. Domain adaptation is achieved through Transfer Learning with DANN, boosting cross-domain detection accuracy to over 90%, a 15-20% improvement over baseline methods. Additionally, an IIPMM using RNN s predicts zero-day attacks with 85-90% accuracy, reducing successful exploits by 20-25%. This framework enhances real-time cybersecurity with adaptive and predictive capabilities, improving network resilience across various environments.