Zero -Day Attack Detection in ۵G Slicing Using Lightweight Reinforcement Learning

The rise of ۵G network slicing has introduced new security vulnerabilities, particularly to Zero -Day attacks, due to its dynamic and virtualized architecture. Traditional detection methods lack the adaptability and efficiency needed to respond to these emerging threats in real time. This study addresses the critical gap in detecting Zero -Day attacks within ۵G slices by proposing a lightweight reinforcement learning (RL) model capable of accurate and adaptive threat detection under constrained computational resources. A modified Q -Learning algorithm was implemented in a simulated ۵G environment, with carefully selected traffic features used for real -time analysis. The model was trained and evaluated using hybrid datasets simulating normal and attack traffic across multiple slicing scenarios. Experimental results show that the model achieved a ۷۲.۴۹ detection accuracy, a false positive rate of %..۹ , and operated with only ۴%۹ CPU usage and a ..۴-second average detection time. These findings confirm that the lightweight RL model effectively balances performance and efficiency, making it suitable for deployment in edge -based ۵G security systems. This research demonstrates that lightweight RL can be a practical and scalable solution for Zero -Day detection, offering significant implications for securing next -generation communication infrastructures.

کلیدواژه ها:

Zero -Day Attack Detection ، ۵G Network Slicing ، Cybersecurity in Edge Computing