Research on intrusion detection in industrial networks based on deep reinforcement learning
In order to effectively identify novel attacks in industrial network environments that are combined by multiple pieces of anomalous data,we propose a deep reinforcement learning-based fusion model DQN-LSTM based on deep reinforcement learning.The model combines spatial and temporal features of traffic data to unfolding anomaly detection .Experiments are conducted on the publicly available industrial control network natural gas plant dataset,and the model in this paper compares favorably with SVM,CNN,and LSTM in terms of accuracy and F1 value.Compared with SVM,CNN, LSTM,DQN and other methods,the model in this paper has better comprehensive performance.
industrial control systemflow anomaly detectiondeep reinforcement learningDQNLSTM
万方数据
维普
