The Method Of The Complex Computer Network Security Assement Based On GABP Agorithem

With the wide use of computer network, the network security becomes increasingly important. At present, the most frequently used network security evaluating methods are risk evaluating method and leak scanning tool software to test and evaluate the security degree of the system. However, network is a complicated system engineering concerning many fields including computer, communications, physics, math, biology, administration and society, etc. Thus the present security evaluating methods still have problems of bad maneuverability, narrow applicability and interference of artificial factors.The title of this dissertation is "The Security Evaluating Methods of Complex Computer Network Based on GABP Arithmetic". The paper mainly contains contents of two aspects.Firstly, based on the opinion of complexity, the author analyzed the basiccharacters and describing indexes of a special complex system——complex network,summarized its research objects and the present research achievements. On the basis, the author further unveiled the main characters of the complicated networks revealed by the computer networks, including openness, avalanche, hierarchy, mutation and uncertainness. Since all the characters are close related to the nonlinearness of the network, the nonlinear security evaluating theory is introduced; the authors made nonlinear analysis on the computer network security evaluation, and put forward the thought of using neural network in network security evaluation.Secondly, using complexity opinion, the security problems of Internet complex network were analyzed. The authors used the comparatively mature evaluating theory of Delphi to carefully screen out the security evaluating indexes and established a comprehensive network security evaluating index system. On this basis we first used the BP arithmetic artificial neural network and founded the comprehensive evaluating model. After analyzing the disadvantages of the BP neural network and the possible improving methods, we improved the model by GA, and carefully analyzed and designed the key parts: the coding manners and flexibility function. The neural network forecasting results on specific samples were detailed compared and tested, and finally proved the advantages of this method. The network security model established in this program can provide some theory reference for complicated science research and it will also help to further realize the network security problem generally.The imitating result indicated that it is easy and feasible of using the model established in this dissertation to evaluate network security. It can eliminate the interference of artificial factors and quickly get the correct evaluating results. This research work provides new thoughts and methods to make comprehensive evaluation to the computer network security, and especially will be worthy referring for future network security problem forecast and control.