Deep Reinforcement Learning based Ensemble Model for Intrusion Detection System

Powered by advancements in information and In-ternet technologies, there has been a rapid development network-based applications. Meanwhile, it is recognized that more attention needs to be paid the issue of cybersecurity. The security network environment plays vital role stable functioning society. Cybersecurity research become active lately. Researchers have proposed several approaches protect network. Among them, broadly practised approach intrusion detection system (IDS). This work suggested potential value reinforcement learning building systems at packet-level. A novel embedding proposed, namely image embedding, encode traffics. Utilizing en-coding raw traffic, which are difficult tackle machine models, can converted images. Thus, experiments applied convolutional neural networks. In addition, packets embedded images arranged time order. this way, integrate flow statistics with packet convert tasks image-associated tasks. experiment, Deep Q-Learning algorithm was selected for ensemble 1D-CNN CNN designed training module an interaction module. Experiments results indicate RL-image-based attain high performance on DDoS traffic provided DDoS2019 outperforms other traditional deep approaches.