Adaptive CyberDefense Against Advanced Persistent Threats Using Game Theory and Reinforcement Learning

Abstract

Advanced Persistent Threats (APTs) are a great challenge to the world of Cybersecurity. The paper makes a contribution to the world of Cybersecurity by developing an adaptive defense system using Game Theory and Reinforcement Learning. In the endeavor to identify anomalies and perform IP blocking and firewall activation, the system proposed monitors real-time data points such as CPU usage, network traffic, and failed login attempts. The system uses a Q-learning engine to optimize its decision-making process, resulting in faster and more accurate responses. The proposed adaptive and learning-driven approach outperforms the existing static security system against emerging APTs, as shown in the experiments.