论文部分内容阅读
在入侵检测中使用单个的支持向量机容易因“单点失效”而危害系统安全.提出一种基于支持向量机集成的方法来进行入侵检测.它采用负相关学习技术,在误差项中使用相关性惩罚因子使得生成的分类器有更好的多样性和精度;算法采用进化策略来自动地确定个体支持向量机的超参数,避免了需要了解问题的先验知识;最后,采用集成技术来组合个体支持向量机的检测结果.仿真实验表明这一方法有更好的检测性能,并且这种分布式并行检测方法有利于增加入侵检测系统的鲁棒性.
In the intrusion detection using a single support vector machine prone to “single point of failure” and endanger the safety of the system.A new method based on support vector machine integration for intrusion detection.It uses negative correlation learning techniques, in the error term The correlation penalty factor makes the generated classifier have better diversity and accuracy. The algorithm adopts evolutionary strategy to automatically determine the parameters of individual SVM, which avoids the need to understand the priori knowledge of the problem. Finally, using the integrated technology To test the results of individual SVM.Experimental results show that this method has better detection performance and this distributed parallel detection method is beneficial to increase the robustness of intrusion detection system.