Deep Reinforcement Learning in Automated Network Vulnerability Detection

Deep Reinforcement Learning in Automated Network Vulnerability Detection

연구 분야: Strategies

학회: 2023 International Conference on Recent Advances in Science and Engineering Technology (ICRASET)

In today's rapidly changing digital environment, identifying security holes in a network is an essential part of maintaining online safety. This research suggests a novel strategy for improving automated network vulnerability identification by making use of Deep Reinforcement Learning (DRL). Our approach combines cutting-edge algorithms to provide a versatile and responsive system; these algorithms include Deep Q-Network (DQN), Proximal Policy Optimization (PPO), and Dueling Deep Q-Network (DDQN). Using six different methods—including signature-based detection, heuristic-based detection, machine learning-based detection, intrusion detection systems (IDS), packet sniffers, and vulnerability scanners—we evaluate how well the suggested method performs in comparison. The detection rate, the number of false positives and negatives, the precision, the recall, the speed of execution, and the capacity to scale are only few of the major performance parameters we use in our analysis. Across a variety of metrics, the results show that the suggested approach performs better than the state of the art. It has an incredibly low false positive (3%) and false negative (2%) rates on top of its 95% detection accuracy. Both its accuracy and efficiency in finding security flaws are demonstrated by its high precision (94%) and recall (96%). The suggested technique also has a low execution time of 25 ms and a good scalability rating of 5, showing it can scale to deal with increasing network complexity. In conclusion, our study demonstrates the use of DRL in automating the process of detecting security holes in networks. The suggested method is an excellent tool for improving network security since it is more accurate and efficient than conventional methods and can be adapted to deal with new threats. By incorporating sophisticated DRL algorithms, a resilient and flexible defensive system may be created, one that can keep up with the ever-changing threats in the cyber world. Show More