基于改进的可变半径检测器实值否定选择算法

Based on Extended Real-valued Negative Selection Algorithm with Variable-sized Detectors

在基于人工免疫的入侵检测研究领域,一般都是先随机产生字符串,然后应用否定选择来生成检测器.这种方法生成检测器的检测率低,而且生成的检测器集数目庞大.本文提出了一种改进的可变半径检测器的实值否定选择算法,由于非自体样本中存在着关于非自体空间的信息,通过应用非自体样本初始化基因库,采用基因库进化策略更新基因库,从而生成了更有效的检测器集.通过实验证明,该方法是有效的,在不影响误报率的情况下提高了检测率.

It is to generate detectors using random strings and then applying negative selection in intrusion detection system based on artificial immune system. The detection rate of the detector set is poor and the detector collection number is huge, by this method. This paper proposed one kind of improvement real-valued negative selection algorithm with variable-sized detectors, there is information about non-self space in non-self samples, so use gene library which is initialized by non-self samples to generate detectors, and apply gene library evolution strategy to update gene library, thus produces more effective detector sets. It is proved by experiments that this method is effective to improve the detection rate without affecting the rate of false alarm.