Abstract: This paper proposes the dynamic pruned binary tree multi-class SVM algorithm against the shortcomings of long training time and slow decision speed in multi-class SVM algorithms.It can reduce the number of support vector and training time effectively.For proofing the effectiveness of algorithm，the paper uses KDD99 data set to evaluate the intrusion detection model applying the preseuted algorithm，and compared with 1-v-r and 1-v-1 algorithm.The results show that the presented algorithm is effective and efficient.

Key words: intrusion detection, multi-class Support Vector Machine, pruned binary tree, kernel function

摘要： 针对现有多分类支持向量机算法所存在的训练时间长、决策速度慢等问题，提出了一种动态剪枝二叉树多类支持向量机算法，该算法能够有效减少支持向量的个数，从而减少训练时间。为了验证算法的有效性，该文使用KDD99数据集对应用该算法的入侵检测模型进行评测，并且将实验结果同1-v-r算法以及1-v-1算法进行了比较。实验结果表明，提出的算法是高效可行的。

关键词: 入侵检测, 多类支持向量机, 剪枝二叉树, 核函数