부산대학교 도서관

The application of information technology has become widespread and popular, resulting in a growing diversity and complexity of network attacks. To address this serious challenge, an effective automatic protection system is urgently needed. This study proposes a novel intrusion detection system (IDS) that utilizes deep reinforcement learning (DRL). Data imbalance problems in the training dataset are addressed by applying the synthetic minority over-sampling technique (SMOTE) and Gaussian Mixture Model (GMM). The agent's strategy is improved through continuous DRL operations, resulting in 78% accuracy on the UNSW-NB15 test dataset. Moreover, the proposed DRL-based IDS can maintain the agent's learning status during its operation.