結合網絡入侵和主機入侵方面的檢測能力，構建了基于智能體的分布式入侵檢測系統的體系結構模型。重點討論了神經網絡入侵檢測算法。針對傳統的BP網絡在入侵檢測應用中學習收斂時間和性能上的不足，提出了變速度回歸神經網絡(采用了批處理技術和動量方法)檢測算法，通過對網絡數據集的測試表明，該算法較傳統BP網絡，其學習訓練次數大大降低，學習能力顯著提高。
關 鍵 詞 入侵檢測系統; 智能體; 分布式入侵檢測; 入侵檢測算法; BP神經網絡

Abstract This paper designs an Agent-based Distributed Intrusion Detection System. The DIDS system combines host-based intrusion detection and network-based intrusion detection functions. It can be used to protect large area network and have relatively good expansibility. This paper also discusses the implementation of BP neural network in intrusion detection. Because of the traditional BP NN’s weakness in learning time and performance, variable learning rate BP regression neural network (batch and momentum techniques are also used) are designed. The algorithm has been tested on a network data set. The result showed that it had much better performance than traditional BP NN.
Key words intrusion detection system; agent; distributed intrusion detection system; intrusion detection algorithm; back propagation neural network