The Research Of Intelligent Intrusion Detection System

This dissertation aims to boost the performance and efficiency of intrusion detection system (IDS) and build up a multilevel intrusion detection model by integrating pattern match and protocol analysis with multi-agent system and parallel genetic algorithm. The dissertation depicts working procedure and module structure in pretreatment system and intelligent intrusion analysis system in details. In the whole system structure, the main purpose of pretreatment system is to filter network data packets quickly and accurately to light the burden of intelligent intrusion analysis system. The intelligent intrusion detection system takes advantage of multi-agent system and parallel genetic algorithm to give a further analysis which aims to improve the analysis ability of intrusion detection system in distributed network and parallel computing environment. When refers to pretreatment system, the dissertation mainly discusses the core technique of data analysis module as well as data classification module. In data analysis module section, the dissertation illustrates the flaws of traditional fast pattern match algorithms and ameliorates them by reducing matching times in specific conditions. In data classification module, the dissertation uses protocol analysis technique which classifies network data packets into distinct classes to call different analysis modules. This approach mentioned above optimizes the searching procedure by using local searching instead of global searching in the whole intrusion matching database. When comes to intelligent intrusion analysis system, the dissertation discusses the system structure of multi-agent system in details and analyzes the work procedure of parallel genetic algorithm. In multi-agent system section, the dissertation introduces a multilevel agent model to enhance the system stability. In parallel genetic algorithm section, the dissertation designs a guided individual migration metric to avoid random migration between different sub-population and uses individual lifecycle function to control the whole population of parallel genetic algorithm.