【摘要】 生物体的内分泌系统是一个高度进化的智能系统,通过激素调节着生物体的神经、免疫系统。受其启发而得到的人工内分泌系统具有强大的调控机制,将其内分泌激素用来调节人工免疫网络的抗体种群进化过程,利用亲合度函数动态调节抗体的克隆规模和网络压缩的规模,充分发挥优秀个体的先进特性来刺激亲合度成熟,并能动态调控种群规模,实现自适应、智能化的网络学习,尤其当样本集边界模糊以及存在噪声样本时,该网络依然可以通过自适应调节有效聚类。最终进化出一个小规模网络来映射原始入侵检测数据集的内在结构。最后,利用图论中的最小生成树对网络结构进行分析,获得描述正常和异常行为的数据特征,得到入侵检测系统的正常模型,由此构建出入侵检测系统。通过在KDDCUP数据集的对比仿真实验,验证了该系统的有效性和可行性,以及对未知攻击的检测能力。更多还原

【Abstract】 The endocrine system of organisms is a highly evolved intelligent system, which regulates the nervous and immune systems of organisms through hormones. The artificial endocrine system inspired by it has powerful regulation mechanism. Its endocrine hormones are used to regulate the evolution process of antibody population in artificial immune network. Affinity function is used to dynamically adjust the clone size and network compression size of antibody, and give full play to the advanced characteristics of excellent individuals to stimulate affinity maturity, and dynamically adjust the population size to achieve self-adaptation. The adaptive and intelligent network learning, especially when the boundary of the sample set is blurred and there are noisy samples, can still be effectively clustered by adaptive adjustment. Finally, a small-scale network was evolved to map the intrinsic structure of the original intrusion detection data set. Finally, the minimum spanning tree in graph theory is used to analyze the network structure, and the data characteristics describing normal and abnormal behavior are obtained, and the normal model of intrusion detection system is obtained, thus the intrusion detection system is constructed. The validity and feasibility of the system and the ability to detect unknown attacks are verified by the comparative simulation experiments on KDD CUP datasets.更多还原