基于主机的入侵检测是保障计算机网络信息安全的重要手段之一 .该文在介绍击键特性及其识别算法的基础上 ,根据大量试验结果 ,确定将具体键的击键序列作为信息源 ,提出了一种改进的贝叶斯统计方法 ,实现了主机入侵检测 .该方法既能独立于用户名和口令之外对用户进行身份认证 ,又能动态监控用户击键的整个过程 .文中对系统实现的关键问题和检测结果进行了详细的分析和讨论 . Host-based intrusion detection is one of the important means to ensure the security of computer network information.Based on the introduction of the keystroke characteristics and its recognition algorithm, based on a large number of test results, it is determined that the keystrokes of specific keys are used as the information source, An improved Bayesian statistical method is proposed to implement host intrusion detection.The method not only authenticates users independently from user name and password but also dynamically monitors the whole process of user keystrokes.In this paper, The key issues and test results were analyzed and discussed in detail.