Abstract: Due to low intrusion detection precision and poor stability of test result，the minimax probability machine and constructive kernel covering apply to Intrusion Detection System（IDS） algorithm is presented.Firstly，this paper uses constructive kernel covering lead data covering maps the awaiting sort samples of original space to a high dimensional feature space，makes the samples linear separable.Then through controlling minimal error of the classification probability and expanding two-dimension classification to multi-dimension space.And the multi-dimension data classification problem with different kernel function is solved through mapping data features to high dimension space.The result shows that the proposed algorithm possesses the features of high classification rate and strong stability.

Key words: intrusion detection, constructive kernel covering, classification, minimax probability machine

摘要：

针对传统的入侵检测算法精度低，结果稳定性差的问题，提出了一种基于构造性核函数覆盖聚类和最大化最小概率机器回归方法的入侵检测算法。首先，利用核函数覆盖将原空间的待分类样本映射到一个高维的特征空间中，使得样本变得线性可分；然后通过控制错分率实现分类的最大化，并利用最大最小概率机的高维映射泛化特性，实现了不同核函数下的数据多维分类问题。实验结果证明，该算法具有分类准确率高、稳定性好的特点。

关键词: 入侵检测, 构造性核覆盖, 分类, 最大最小概率机