Peer To Peer Traffic Identification Using Hybrid Cuckoo Search And Particle Swarm Optimization Algorithm

With the development of network, nowadays, the Peer-to-peer (P2P) technologyhas become one of the most important technologies in the Internet. However, a lot ofproblems arise at the same time, such as rampant piracy, low quality and leakage ofsensitive information. In order to make P2P technology provide better service for people,accurate identification of P2P traffic makes great sense for efficient networkmanagement and reasonable utility of network resources.P2P traffic identification is a pattern recognition problem in nature, the accuracy ofthe P2P identification largely depends on selecting the traffic feature and building theclassification model, in order to improve the accuracy of identification, this papermainly includes the following parts:(1) P2P traffic feature selection using hybrid Cuckoo Search and Particle SwarmOptimization (CS-PSO) Approach. In the problem of P2P traffic identification, theaccuracy is low when use a single feature to identify traffic. To improve the accuracyof traffic identification, a variety of features are introduced to identify traffic. However,too many features may also bring the problem of excessive dimension. Although the useof SVM (Support Vector machine, SVM) classifier can overcome the curse ofdimensionality, it will increase the workload of collection sample, and decrease theaccuracy of P2P traffic identification. In this paper, in order to improve the accuracy andcomputational efficiency of P2P traffic identification, a novel method based on hybridCuckoo Search and Particle Swarm Optimization Approach is introduced to select theoptimal feature subset.(2) P2P traffic identification using CS-PSO algorithm and SVM. SVM is amachine learning approach based on the statistical theory, which can find the optimalsolution of the classification results by provided limited information about a smallsample dataset. It avoids the shortcomings of many machine learning approachesrequire large sample data sets and according to specific problems to establishappropriate model by use of nonlinear approaches. As one of the optimal classifiers,support vector machine (SVM) has been successfully in P2Ptraffic identification.However, the performance of SVM is largely dependent on its parameters. The commonused approaches are grid search approach, genetic algorithm (GA), PSO algorithm andso on, but these methods exists some problem. In the paper, in order to improve theaccuracy of P2P traffic identification, a novel method based on hybrid Cuckoo Search and Particle Swarm Optimization Approach is introduced to optimize parameters ofSVM.Finally, the proposed methods of feature selection and SVM parameteroptimization are compared with existing GA algorithm and PSO algorithm, through thetest in UCI machine learning database and P2P data of real campus network, theexperimental results demonstrate that feature selection approach based on CuckooSearch and Particle Swarm Optimization algorithm can obtain excellent feature subset,and the optimized SVM can significantly improve the performance of trafficidentification.