基于深度Q学习的网络入侵检测强化学习方法

A Reinforcement Learning Method for Network Intrusion Detection Based on Deep Q-Learning

提出了一种基于Q学习的网络入侵检测方法,将强化学习与深度前馈神经网络方法相结合,为网络环境提供了持续的自主学习能力.同时,使用自动试错方法检测不同类型的网络入侵,并不断增强其检测能力.此外,还提供了微调深度Q学习模型中涉及的不同超参数的细节,以实现更有效的自适应学习.基于NSL-KDD数据集的大量实验结果表明,提出的深度Q学习模型能效检测不同的网络入侵类型,检测准确率优于其他机器学习方法.

A network intrusion detection method based on Q-learning is proposed,which combines reinforcement learning with deep feedforward neural network method to provide continuous autonomous learning ability for network environment,and automatic trial and error methods are applied to detect different types of network intrusions and continuously enhance their detection capabilities.The details of different hyperparameter involved in the fine-tuning depth Q-learning model are provided to achieve more effective adaptive learning.A large number of experimental results based on the NSL-KDD dataset indicate that the proposed deep Q-learning model is very effective in detecting different types of network intrusions and outperforms other machine learning methods.