关键词: 计算机网络, 信息安全, 入侵检测

Abstract: In order to detect intrusions of high-speed network，the mechanism of intrusion detection is discussed.Intrusion detection is reduced to an inference procedure on an incomplete data set.The concepts of self-similarity degree and completeness degree of the knowledge base are proposed.They are used to control the scale and increasing speed of the knowledge base so that intrusion detection is assured to be proceeded in a limited space.At the same time，information gain is used to make intrusion detection to running in a lower space.The experiments show that the proposed method reduces the calculation load of intrusion detection systems effectively and enhances its real time performance.

Key words: computer network, information security, intrusion detection