Abstract: Security audit is one of the most important techniques to protect computer from misuse and malicious attack.Security audit system uses trail rules to trace the users’ action.Unfortunately，it is difficult for administrator to set the accurate rules for every user because of the lack of the priori knowledge.This paper introduces genetic algorithm-based trail algorithm to solve this kind of intractable problems.Experiments show that the algorithm is usable and the performance of the approach has been improved greatly at the acceptable cost of the accuracy（80%）.It helps on the security audit system to process the increasing number of the audit data and users.

Key words: information security, security audit system, data mining, genetic algorithm

摘要： 安全审计是保护计算机遭受恶意攻击的重要技术之一。安全审计系统利用跟踪策略来实现对用户行为的跟踪。但是由于先验知识的局限，对于管理员来说很难为每个用户制定准确地跟踪策略。介绍了一种基于遗传的审计跟踪算法解决此类问题。实验证明，该算法具有实用性，并能以算法精度上损失可接受的代价（接近最优解80%以上）极大的提高算法的运行时间。使得该方法可以用在大规模多用户的环境。

关键词: 信息安全, 安全审计系统, 数据挖掘, 遗传算法