An AI-Based Cyber Ranges to Strengthen the Cybersecurity of Cyber Physical Systems
Deepa Singh Sisodiya, Ritu Tiwari, Priyank Jain, Yashwant Aditya · Journal of Applied Security Research · 2025
This research demonstrates that integrating artificial intelligence into cyber range platforms significantly enhances cybersecurity readiness for cyber-physical systems by improving threat detection accuracy, accelerating incident response, and enabling adaptive learning. We developed a hybrid AI-powered cyber range architecture combining cloud-based simulation for scalability with emulation-based components for physical system fidelity. The framework leverages LSTM and GRU networks trained on 2.6 billion security events and implemented using Python 3.9 with Keras/TensorFlow, optimized via Adam optimizer (90.0% accuracy vs. 85.9% for ADAMAX). Results revealed three critical advancements: 91.3% classification accuracy in detecting coordinated attacks, identification of 76/107 security vulnerabilities (71% success rate) with 89.47% concept recognition, and a 34% reduction in detection-to-mitigation time compared to conventional cyber ranges. While demonstrating superior performance in controlled environments (90.9% accuracy in patch validation), challenges persist in AI explainability—only 58% of cybersecurity professionals could interpret model decisions, underscoring the need for interpretable machine learning frameworks in operational deployments.