Research On Traffic Anomaly Detection Method Based On Transfer Learning And Improved Genetic Algorithm

People's work and study are more and more inseparable from the network.Network brings convenience to people's life,at the same time,network vulnerability attacks,such as denial of service attacks,sudden access,worms,also threaten people's privacy and property security.Traffic anomaly detection plays an increasingly important role in detecting and preventing potential threats.In the field of traffic anomaly detection,there have been a lot of research results,but there are still some problems,such as low detection rate for unknown attack types and low recognition rate for a few categories.In this paper,some of the problems are studied.Firstly,Traffic anomaly detection based on transfer component analysis: Aiming at the problem that traditional machine learning algorithm has weak ability to identify new network attacks,a traffic anomaly detection method based on transfer component analysis is introduced.Firstly,feature subset is selected to remove redundant features.Secondly,the existing attack knowledge is transferred to the new attack by reducing the distribution distance between the source domain and the target domain.There are seven kinds of transfer tasks in the two kinds of experiments.The experimental results show that this method is better than the traditional machine learning method and the feature-based transfer learning method He Map.Then,Detection of a few abnormal traffic based on improved genetic algorithm:Aiming at the problem that the existing research methods focus on improving the overall accuracy rate but ignore the problem of small sample size,a method of traffic detection based on improved genetic algorithm is studied in this paper.This method uses the arithmetic average of the error rate of normal class and abnormal class to replace the total classification accuracy,so as to improve the genetic algorithm for feature selection.SVM and decision tree with better classification performance are used as classifiers.The experimental results show that compared with the traditional feature selection based on genetic algorithm,the proposed algorithm has certain advantages in the recognition rate of small classes.