激素调节的克隆选择聚类在入侵检测中的应用-《计算机技术与发展》

文章信息/Info

Title:: Application of Clonal Selection Clustering Algorithm in Intrusion Detection Based on Hormone Regulation

Author(s):: BAI Lin; School of Computer Science and Technology,Xi’an University of Posts and Telecommunications, Xi’an 710121,China

Keywords:: artificial endocrine system; hormone regulation; clonal selection; clustering analysis; intrusion detection

摘要:: 受生物体的内分泌系统和免疫系统相互作用、相互影响、协调拮抗的工作机理启发,将人工内分泌系统的激素调节机制作用于抗体种群的进化过程,由当前抗体亲合度动态调节克隆规模,使得优秀个体在进化过程中得以扩增充分发挥其优势,不良个体得到抑制,以此刺激亲合度趋于成熟。采用改进的克隆选择策略指导聚类过程,构造基于激素调节的克隆选择模糊聚类新算法。激素调节下的多种进化算子具有良好的自适应性、自学习性和稳定性,保证了聚类算法全局寻优的速度、精度以及可伸缩性。基于该方法构建了网络入侵检测系统,利用 KDD CUP 99 数据集构造训练集和测试集,并采用具有混合属性的相异性度量函数,仿真实验表明该系统性能理想,并能有效检测出未知攻击。

Abstract:: Inspired by the interaction and coordination of the endocrine system and immune system,the hormonal regulatory mechanism of artificial endocrine system is applied to the evolution of antibody population. The clonal scale is dynamically regulated by the affinity of current antibody. The excellent individual can be amplified and the bad individual is suppressed in the process of evolution,which stimulates the affinity maturation. The improved clone selection strategy is used to guide the clustering process,and a new fuzzy clustering algorithm is proposed based on hormone regulation. Multiple evolutionary operators based on hormonal regulation have better adaptability,self-learning and stability. The global optimization speed,precision and scalability of clustering algorithms are ensured. Based on clustering analysis,a new network intrusion detection system is constructed. The training set and test set are constructed by KDD CUP 99 dataset. And a mixed function dissimilarity function is adopted. The simulation shows that the system has ideal performance and this method can detect the unknown attacks effectively.