Towards Secure-by-Design Systems: A Policy-Guided Framework for AI-Augmented Cyber Defense

In response to the growing complexity and sophistication of cyber threats, this research proposes a policy-guided, AI-augmented cyber defense framework that integrates metaheuristic feature selection, explainable AI, and semantic policy compliance mechanisms. Designed with a secure-by-design philosophy, the system addresses key limitations of traditional and hybrid intrusion detection systems by enforcing policy-aligned decision-making and improving transparency. Empirical evaluation across standard datasets demonstrates the framework’s superior performance, achieving 97.1% accuracy, low false positive rates, and real-time detection capabilities while maintaining full policy compliance. Enhanced explainability using SHAP values improves analyst trust and decision assurance. The results affirm that embedding policy logic and interpretability into AI models not only boosts security performance but also aligns AI-based detection systems with governance and regulatory demands. This study contributes a comprehensive, accountable, and adaptable cyber defense architecture suited for modern threat environments.