Research On Worm-Propagation Prediction Based On Scale-free Network

Each time the eruption of the Internet worm and network virus happened, which usually cause quite serious network paralysis and huge property loss. A large number of studies show that the worm propagation is a Markov process whose spreading is related to the current complicated network state and the network topology structure. In order to achieve good prediction results, the current domestic and foreign study for predicting worm mostly concentrated in studying and Improving Markov chain model, however, the network worm left the opportunities of effective defense fleeting due to its varies kinds of type and fast transmission speed. Therefore, it does not have excellent guidance sence and practical value.This thesis uses the prediction way of “meeting changes with constancy” which applying nonlinear SVR model to forecasting the trend of Internet worm propagation, based on the node characteristics of the scale free network. In general, the real complex network has the characteristics of complex structure, evolving network, diversified connection and the multiple fusion of complexity and so on, which make the form of complex network difficult to grasp, but the only thing that remained unchanged is the feature of "small world" and scale-free. workers for network security and staffs for the operation and maintance of system security can make a more accurate prediction of network worm propagation in the validation of its own network characteristics, and take time for worm attack. This research give the practical guidance to the defense of worm propagation, and provides a new idea for the research of Internet worm virus. The specific research contents are as follows.Firstly. The complex network is the medium for worm propagation. The position and status of each node in the network is described by the analysis of the common complex network evolution model, which contain rule network model, ER random graph model, WS small world network model, BA scale-free network model, and extract the network node characteristics such as the degree distribution, the single source shortest distance, network clustering coefficient, the rich club and closeness central coefficient, etc.Secondly. The worm propagation behavior is the object of this research, the dynamic process of worm propagation model, SI model, SIS model, SIR model are analyzed, and the reliability has proved by simulation in complex networks, which shows that the higher the degree of nodes and the less power law exponent of the network, the more useful to worm propagation.Thirdly. The prediction of Internet worm propagation is the purpose of this research. So the nonlinear SVR algorithm is used to predict the propagation and the results are analyzed and tested. A nonlinear SVR prediction model is established and the trend of worm propagation in the network has been predicted accurately by analyzing and comparing the linear regression model and SVR model. And finally, the superiority of the nonliner SVR model to the high dimensional data is found by the error analysis.