Real Value Negative Selection Algorithm Is Studied And Improved

In recent years, the human rights' dependence on information is increasing, and information security is increasingly concerned on.It is discovered in researches that there is comparability between the problems on intrusion detection field and artificial immune system. Therefore, many of the immune system satisfactory features can be used to solve the problems that happened in computer security field and the research of managing the mechanism of immune system in computer security field is regarded.Forrest proposed the negative selection algorithm based on the principles of self and non-self discrimination in the immune system, and the negative selection algorithm was the first immune-inspired anomaly detection algorithm proposed. To improve the performances of detectors, the real-valued negative selection algorithm was proposed by improving detector generation algorithm based on the negative selection algorithm.Real-valued negative selection algorithm applied an iterative process that updated the position of the detectors driven by two goals: one is moving the detector away from self points, and another is keeping the detectors separated in order to maximize the covering of non-self space. But there are still some problems on real-valued negative selection algorithm, such as needing large amounts of time in generating detectors and setting the position of detectors. Therefore, the main problem to be solved in this paper is how to use the least detectors to cover the most non-self space.The paper analyses current success and researches the application of negative selection algorithm and real-valued negative selection algorithm in intrusion detection by the direction of artificial immunology, including changing the radial of detectors and a complex algorithm based on real-valued negative selection.Then this paper gives ameliorations on two aspects based on real-valued negative selection algorithm. One is the amelioration on calculating the size of the non-self space based on Monte Carlo simulation techniques, and another is the amelioration on antibody distribution. Finally, the results of experiments testify that new algorithm has advantages on the size of non-self space covered by antibody, and the new algorithm has in terms of maximizing the covering of the non-self space.