Research On Identification And Classification Algorithm For Encrypted Traffic In Wireless Networks

With the rapid development of network technology,wireless network is taking up an increasing proportion of the network,which is an important transmission medium at the network edge.Different types of wireless networks such as Wi-Fi,Bluetooth,in-vehicle network,and satellite communication generate a large amount of wireless network traffic in reality.Meanwhile,with the improvement of the security consciousness of the public and the whole society,wireless encrypted traffic has become an important base of the development of the Internet.Though wireless networks and wired networks are different in protocol format and encryption method,the emergence of encrypted traffic has a certain impact on network management.At present,people in the industry usually analyze the statistical characteristics of wireless network encrypted traffic.However,these features are not efficient in classifying and identifying wireless network encrypted traffic,and there are certain difficulties in identifying services of untagged wireless encrypted traffic.In view of the above problems,this paper studies the encrypted traffic of wireless network.Through the data link layer protocol analysis of encrypted traffic of wireless network,a feature set for different wireless network encrypted traffic is constructed,and the feature selection algorithm is studied for the feature set.Based on the optimal feature subset,a related algorithm is proposed for the identification of untagged wireless network encrypted traffic services to realize wireless encrypted traffic service identification.Based on the research work,a wireless network encrypted traffic service identification and classification system is designed and implemented.The main contributions in this paper are as follows:(1)Aiming at the clustering problem of unlabeled datasets,a feature construction method of the data link layer(WNFC)is proposed by taking the obtained wireless network encrypted traffic as the research object.The communication principle and the MAC frame format of the wireless encrypted traffic construct the wireless network encrypted traffic feature set.(2)Aiming at the problem that the constructed feature set has high feature dimension and contains redundant features and irrelevant features,a feature selection algorithm(GA-NB)is proposed,which uses Naive Bayes as the fitness and optimizes the genetic algorithm.The GA-NB feature selection algorithm uses the classification accuracy of the Naive Bayes model as the fitness function,evaluates the similarity between the feature and the search problem,and selects the optimal feature subset for wireless network encryption traffic.The experimental results show that the feature selection algorithm proposed in this paper can effectively select the optimal feature subset in the feature selection of wireless network encrypted traffic,the feature set optimization rate achieves54.5%,and the algorithm has good stability.(3)Aiming at the problem of business identification of unlabeled data,a clustering optimization(OPK)algorithm based on K-means and a shortest distance matching(FWSDM)algorithm based on feature weighting are proposed respectively.The OPK algorithm realizes the clustering operation of unlabeled data traffic by selectively selecting the cluster center points.The FWSDM algorithm calculates the distance between each piece of data and the center point of different business data sets based on the feature weight value,finds the shortest distance through comparison,and matches the data with the business type.The experimental results show that the clustering accuracy of the OPK algorithm can reach 94.45%,and the business matching accuracy of the FWSDM algorithm to unlabeled data can reach 96.89%.(4)On the basis of the above work,a wireless network encrypted traffic business classification and identification system is designed and developed,which realizes the related functions of different modules through the framework design,process design,and function design.This paper also tests the performance,running state,and abnormal state of the system.The test results show that the system runs normally and has a friendly interaction with users.