基于人工免疫的新型检测器生成模型

继承了人工免疫系统的思想，研究了KIM和BENTLEY的克隆选择算法，提出了一种适用于入侵检测的新的检测器生成模型。其核心在于两个新的算法：一是为了提高检测器的多样性及适应度水平，提出了基于相似性和适应度相结合的概率选择算法，并给出了此类概率选择的一般形式，理论分析了算法中的权重参数α。二是在产生子代检测器时，为了使得父代的优良基因能最大程度地遗传给子代，防止交叉变异中的退化现象，提出了检测器有效因子的概念和使用有效因子进行保优的策略。通过仿真实验证明合适选择α参数以及有效因子的长度阈值Neg，能使该模型具有很好的多样性和自适应性，呈现出较高的“非我”检测率和低的误检率。

A new model for detector generation based on artificial immune system

Based on the research of clonal election algorithm of KIM and BENTLEY, a new model for detector generation, which inherited the idea from artificial immune system (AIS), was proposed. This model worked with two key newly developed algorithms. Firstly, in order to improve the diversity and adaptability of detectors, a probability selection algorithm based both on similarity and adaptability was proposed, and the common form of this kind of probability selection was given. The authority factor above in this algorithm was analyzed theoretically. Secondly, in order to pass the best gene of father detectors to the children as well as possible and prevent degradation in process of mutation and crossover, the concept of effective gene and the strategy to keep excellence were proposed. Simulation results suggest how to choose appropriate parameter α and threshold Neg of length of effective gene. The ideal choice allows the model to achieve high level of diversity and adaptability, and have good non self detection rate with a very low rate of self detection.