基于免疫多Agent的网络监控系统模型研究

Study on Agent Immune Network Monitoring System Model

研究入侵检测问题,针对网络免疫系统检测器训练速度慢、网络系统自适应性差和阈值量化等问题,提出了免疫多A-gent的网络监控系统模型。在模型中,首先以抗体激活阈值为度量对网络事务集进行自体、非自体分类和网络成熟检测器的生成;然后对成熟检测器通过克隆优选策略和检测器影响权重函数进行分布式网络系统的成熟检测器筛选与优化,依据免疫系统的初次耐受应答生成能够对非自体抗原进行识别的记忆检测器;最后利用记忆检测器对实时获取的网络系统窗口数据进行抗原识别。仿真结果表明,提出的算法具有较好的检测率和较低的误测率,同时有效的降低了检测器的训练时间。

In order to solver the problem of network immune system detector slow training speed, network system adaptability difference and threshold value quantification, an agent immune network monitoring system model was pro- posed. In this model, firstly, the network transaction was sorted as antilogous and non-antilogous, and the network mature detector was generated with the antibody activation threshold as a measurement. Then, through cloning optimi- zation strategy and detector influence weighting function, the mature detector was screened and optimized. According to the immune system＇s first tolerance response, the memory detector was generated to recognize non-autoantibody, Finally, the memory detector was used to recognize antigen form the network system window data obtained in real- time. The simulation results show that this algorithm has higher detection rate and lower false rate of measurement, and at the same time, effectively reduce the training time of the detector.