基于并行PSVM(proximal support vector machine)分类法,利用ε-支持向量与原数据集等价的特点,将PSVM和cascade SVM模型高效结合,加速训练入侵数据集.提出一种新的PSVM增量学习方法,它能快捷更新分类器.通过大量基于著名的KDD CUP1999数据集实验,研究表明,该算法相对其他SVM方法,在保证较高检测率和较低误报率的同时,其训练时间降低80%,且能通过增量学习新数据集来有效更新分类器. Based on the parallel support vector machine (PSVM) classification, the PSVM and cascade SVM models are efficiently combined to accelerate the training of intrusion datasets by using the feature that ε-support vector is equivalent to the original dataset.A new PSVM incremental learning Method, which can update the classifier quickly.Experiments show that this algorithm reduces the training time by 80% %, And effectively update the classifier by incrementally learning a new data set.