Research On Detector Generating Algorithm Based On Negative Selection

The research on artificial immune system aims to extract the special information-processing mechanism contained in biology immune system and design corresponding models and algorithms, and solve all kinds of complicated problems finally. Artificial Immune, as an interdisciplinary subject of life science and computer science, is a new field of intelligent computation. It has become a prevalent subject.The basic function of biological immune system is to distinguish "self" from "non-self", and get "non-self" away. Biological immune system, as a self-adaptive, self-learning, self-organizing and parallel- processing system, has several functional characters such as immunity recognition, immunity memory, immunity regulation and immunity surveillance. By intensive research on kinds of information-processing mechanism contained in biology immune system, designing effective intrusion detection model and algorithm, it has great significance on the setting up of new theory and new method based on biology immune system, which would improve the internet security.The essay is mainly about the history of development of intrusion detection system; the introduction of the function constitution classification of intrusion system; explore and research on learning and detection mechanism of biology immune system and application of artificial immune system in anomaly detection. The essay also improves the present detector set generation algorithm based on the analysis of the application mechanism and defect of negative selection---one of the core algorithms in artificial immune system and analyzes the performance of generated detector set from several aspects by tests. Concretely speaking, the main researches of this essay are as follows:①Through analyzing the process of negative selection algorithm and the generating mechanism of detector set, the essay analyzes that the detector set generated from negative selection algorithm causes problems of a fixed matching threshold , large amounts of black holes, detectors match each other, information redundancy and decrease of space coverage ratio.②Based on the research of the application scope and defect of present r-contiguous matching regulation and r-chunk matching regulation, the essay uses the concept of fuzzy thinking to define local similarity degree and diversity degree. The essay compares the detector set scope and distribution generated by different matching thresholds and theoretically analyses the relationship between matching threshold change, black hole amount and space coverage ratio.③The essay also designs threshold change strategy, improves traditional negative selection algorithm, puts matching threshold which corresponds with the detector into the generated detector set, enhances the space coverage of the detector set and decreases the amount of black holes. In addition, the improved algorithm gets rid of the detectors that match each other, eliminates information redundancy and can detects a wider scope of"not-self"behavior while generating the same detector set scope.④At last, the essay compares and analyses the improved algorithm and other calculations in the aspects of black hole number, space coverage ratio and failure rate to verify the effectiveness of the improved algorithm.