An Improved Feature Selection Algorithms And Anti-DDoS Analysis

Abstract:With the growing of network’scale,the network’s environment has been becoming more and more complex,the number of network intrusion is also increasing,it do harm to the network security and the vital interests of the people seriously,and it seriously hinder the development of the network too. With further increase in network features,every algorithm has its own limitations.At the same time, it is almost impossible to find a reliable method to deal with network DDoS attacks.It’s urgent to build a better intrusion detection system model and find a more efficient feature selection algorithm.In this paper.We have summarized and inheritted the traditional intrusion detection systems strategy by conducting in-depth analysis and research in some excellent traditional intrusion detection system and feature selection algorithm.We have proposed an improved model based on reputation collaborative.It have developed the second-level cache and buffer queue.In order to access the data stream from the serial-to-parallel conversion,the program was done by the network server downlink packet depositing correspondly the cached queue on the basis of the IP address and extracting with a sliding window,and used buffer overflow management mechanism to block attack traffic in a timely manner.At the same time,the packet that can’t be process immediately can be storing in the second-level cache.In this program,we have proposed an improved information gain feature selection algorithm.In this algorithm,we will calculate the correlation value of each feature,and then, add features to feature subset follow the slow-start strategy,at last,using the evaluation function to evaluate the features subset to obtain a set of optimal feature set as soon as possible. The simulation test has compared the improved model and the original model.The result showed that the improved model can compensate effectively for the deficiencies of the original model in the detection defense and the improved feature selection algorithm is more efficient.