A Detector Generation Algorithm Based On Co-evolution

With the development of computer technology and the popularity of network applications, the traditional security measures based on passive defense are unable to adapt to today's changing network environment. How to effectively protect important information in the computer and how to create a secure network environment for uses have become a key issue that must be considered and resolved for current network security.Immune-based intrusion detection technology in its treatment of active defense and intelligent technology, and characteristics of its distributed protection has become a hot research topic in network security field. Through intrusion detection technology, the system can still maintain its safety and maneuverability and continue to provide critical services when suffering from external attack.The detection rate based on immune intrusion detection system is mainly decided by the detector non-self space coverage, the detector is mainly used traditional genetic algorithm to generate too much emphasis on the individual struggle within species, ignoring the linkages between the individual has certain one-sidedness, the generated detection rate of the detector is much lower. To solve this problem, we propose a detector generation algorithm based on co-evolution. The algorithm divided the population that need to evolve into several sub-populations through simulating co-evolution mechanism in nature. The single species evolved by genetic algorithm, and the various sub-populations interconnected by mutual influence on behalf of individuals, in order to achieve co-evolution of the system, which breakthrough the limitation of independent evolution of the population , and finally improved the detection rate of the detector.For the population separating problem existing in the co-evolution algorithm, and in order to maintain the tight joint between populations is not destroyed, we propose a population partitioning algorithm based on tight coupling of probability. The algorithm determines the division of populations by calculating the close relationship between the individual threshold, which not only can maintain the tight joint between populations, but also can maintain the population diversity.This thesis presents corresponding algorithm for detector generation in immune intrusion detection system and population division in co-evolution, and makes analysis and experiments for the algorithm.