Research On Network Security Situation Awareness Based On Neutral Network

According to the increasingly serious situation of network security, traditional security defense system can provide lots of alarm information, but is weak in coordinating responses to attack in real-time. Network security situation awareness, as a kind of active security defense system, can not only dynamically reflect the condition of network operation, but can also predict the trend of development. Based on the discussion of the theory about network security situation awareness, an index system of network security situation assessment is built,as well as methods of security situation assessment and prediction based on neural network are proposed.1. An index system of network security situation assessment is built. In this system,network security situation assessment levels are defined to describe the network security situation from four independent properties: menace, vulnerability, tolerant and stability, also,methods to get situation elements and computing indexes are given. The problem about singleness of index data source is solved efficiently, which provides full support for security situation assessment and prediction.2. Present a method based on BP Neural Network to estimate network security situation assessment. The method that abstracts input layer, hidden layer and output layer from assess model, which, then will be formed a comprehensive evaluation model with non-linear mapping ability and strong adaption via the training of BP algorithm, which can solve the problem on uncertainty and fuzziness between the factors and assessment of network security situation.3. Present a method based on RBF Neural Network to forecast network security situation.The method that can find out non-linear mapping relationship between situational values of previous N time-frames and later N time-frames by means of RBF Neural Network according to the network situational value's characteristics that are non-linear time series and optimize the network parameters by using adaptive genetic algorithm, which is more appropriate for the real network environment since it makes up the traditional method's defects on forecasting speed and precision.4. Prove the feasibility and effectiveness of given method on evaluation and prediction.The simulation experiment proves that the method for network security situation assessment based on Neural Network can get more accurate situational assessment and prediction results.