Chunzhi Wang

Study of Double-Characteristics-Based SVM Method for P2P Traffic Identification

In this paper, an SVM (Support Vector Machines)-based P2P (Peer-to-peer) traffic identification algorithm is presented. It could capture traffic information online, training-offline and categories online. The SVM algorithm uses double characteristics IP and IP-Port to identify P2P traffic by means of different traffic features separately. From results of experiments, we proved that choosing the appropriate traffic features, kernel option, configuration parameters, and punish modulus to the RBF kernel function of SVM algorithm are effective to identify P2P traffic.

The Study of DPI Identification Technology Based on Sampling

DPI identification technology is a main technique of P2P data stream identification technology; the advantages of it are higher accuracy and the ability of distinguishing accurately which type of P2P applications. But it is difficult to apply in highspeed networks due to its slow execution speed. To solve this problem, this paper suggests six sampling strategies: invariable random sampling, invariable mechanical sampling, time random sampling, time mechanical sampling, speed random sampling mode and speed mechanical sampling. Discussing and testing how these strategies affect the DPI identification system in different network environments. Test conclusions shows that six sampling modes reduce the burden on the system in different degrees and maintain a high identification ratio of P2P data stream. Keywords—Sampling; Deep Packet Inspection; Peer-to-Peer; Identification

Improved K-means algorithm based on hybrid rice optimization algorithm

Clustering analysis is an active research branch in the area of data mining due to its simplicity and rapidity. However, K-means algorithm has the shortcomings of heavily depending on the initial clustering center and easily falls into local optimum. In this paper, we consider a deep research on K-means algorithm of optimization. We put forward the first selected initial clustering center of K-means algorithm, toward this end, a novel hybrid algorithm based on K-means algorithm and Hybrid Rice Optimization Algorithm were proposed to rapidly find the optimal cluster centers and avoid getting into local optimum. Experimental results show that the proposed clustering algorithm outperforms other similar algorithms.

Algorithm Comparison of P2P Traffic Identification Based on Deep Packet Inspection

Three classic multi-string matching algorithms such as AC, Wu-Manber and SBOM are researched and implemented in P2P traffic identification system based on Deep Packet Inspection in this paper. In order to adapt to the DPI technology, the code and parameter of these matching algorithm be make suitable adjustment. Three algorithms are used in DPI scanning under the same testing environment and the pretreatment time, analysis speed, degree of accuracy from this testing were compared. Based on these experimental data, the superiority-inferiority of these three algorithms in different conditions is discussed and their relative scopes of application are summarized.

An ACO-based Link Load-Balancing Algorithm in SDN

Software Defined Networking is a novel network architecture, which separates data and control plane by OpenFlow. The feature of centralized control to acquire and allocate of global network resource. So, the link load-balancing of SDN is not as difficult as the traditional network. This paper proposes a link load-balancing algorithm based on Ant Colony Optimization(LLBACO). The algorithm uses the search rule of ACO and takes link load, delay and pack-loss as impact factors that ants select next node. For maintaining link load-balancing and reducing end-to-end transmission delay, the widest and shortest path in the all paths can be gained by ants. The results of simulations show that LLBACO can balance the link load of network effectively, improve the Quality of Service(QoS) and decrease network overhead, compared with existing algorithm.

The research of recommendation system based on Hadoop cloud platform

This paper integrates hadoop with recommendation system. Recommendation system is a system which makes predictions of user preferences. In this paper, we introduces a improved hybrid recommended algorithm and combine MapReduce programming which is used on Hadoop plaform. Through practice, the improved algorithm can accurately acquiring user preferences, provide the user with recommendation when the user browses the web page. Finally, Hadoop can meet the demand of massive data processing, and achieve a highly performance system by making use of the data reprocessing method.

Some ideas on the computer hardware curriculum construction under the internet of things era

In order to improve the computer hardware courses teaching quality under the Internet of things era, based on the analysis of the inadequacies of computer-related hardware course teaching, We put the computer hardware curriculum group construction as the target, discuss on system structural adjustment and optimize of computer hardware course group, and give some suggestions for computer hardware course teaching.

A SVM Method for P2P Traffic Identification based on Multiple Traffic Mode

Support Vector Machines (SVM) algorithms are one of the algorithm s currently applied in Deep Traffic Inspection (DFI) technologies. This paper realizes online real-time traffic information detection, provides a P2P traffic identification system that supports online SVM analysis and offline SVM training function, and demonstrates the thinking of different identification for IP data traffic and IP-Port data traffic. This paper designs different combinations of traffic features for IP data traffic and IP-Port data traffic, analyzes the effectiveness and exactness of these combinations from various function criteria, and based on a lot of experiments, obtains a best SVM kernel function and a combination of parameters that match es the very combination of traffic features.

A Peer-to-Peer Game Model using Punishment Strategies

Recent years, with the rapid development of P2P networks, the security problem of these networks has become increasingly obvious. According to the behavior of selfishness and betrayal for nodes in P2P networks, this paper analyzes and contrasts the benefits of these nodes, and presents the P2P game model with the penalty factor. The reasonable analysis and simulation by gambit prove that, addition of the penalty factor has certain constraints on the betray nodes and promotes the active cooperation of these nodes, thus improves the security status of P2P networks.

Peer to peer traffic identification using support vector machine and bat-inspired optimization algorithm

Nowadays, Peer-to-Peer computing technology (P2P) is widely used on Internet, which has brought great challenges to effective management of the network. As a result, it is very important to recognize P2P applications as to maintain network. In essence, to identify traffic of P2P is a problem belongs to pattern recognition. As one of the optimal classifiers, support vector machine (SVM) has special advantages with avoiding local optimum, overcoming dimension disaster, resolving small samples and high dimension for P2P classification problems. However, the performance of SVM is largely dependent on its parameters and the traditional tuning methods are inefficient. Therefore, in the paper the bat algorithm is proposed to seek the optimal parameters for SVM. In the end, experimental results display that the proposed method outperforms SVM optimized by genetic algorithm, particle swarm optimization algorithm, which can effectively improve the accuracy of P2P network traffic identification.