Research On Key Security Technology Of Command Automation Network Based On Neural Network

As a multiplier of military strength, command automation system affects the process offighting in the complicated and changing battle environment. The command automationnetwork, as the basis of implementing the functionality of command automation system,undertakes the transmission task of military information and plays a vital role in the processof effective fighting. Therefore, the security of command automation network is particularlyimportant. And defense measures such as firewall, intrusion detection, etc. have been takenaccording to various threats.In view of defects of traditional BP(Back Propagation) algorithmused in the firewall and intrusion detection, such as slower convergence speed, easy to beentrapped in local minimum and so on. algorithm is improved. The wavelet neural networkalgorithm and LM(Levenberg-Marquardt) algorithm are applied to forecast of firewall flowand the classification of the intrusion detection. The simulation results can be got from theaspect of convergence speed, prediction error and classification effect and it is effective toapply the improved method to firewall flow and the classification of the intrusion detection.Concrete content is as follows:Firstly, to establish models respectively according to the forecast of firewall flow andintrusion detection. The research according to the characteristics of the military firewall flowfocuses on affecting factors of firewall flow including the current flow data, week and timeetc. The relationship between the input and output is determined to establish the model offirewall flow forecast. As a complement to the firewall, intrusion detection is a kind of activedefense measure. When firewall cannot correctly recognize and prevent some deceived link,data information in the key nodes of command automation network can be collected activelyas input of neural network by intrusion detection. The dataset of KDD CUP99will be used tomake a model to test the safety of current command automation network and calssify theinvasion.Secondly, to respectively optimize the algorithm of firewall flow forecast and intrusiondetection. One of the important application fields of neural network algorithm deals withprediction problems. Acoording to the better effect of the wavelet algorithm applied in theproblem of time series prediction, it is used in the prediction of command automation networkfirewall. Through the simulation analysis and the comparative analysis of the convergencespeed and prediction error, the result is achieved, and another important application field ofneural network algorithm is to solve the classification problem. Aiming at the defects of thetraditional BP algorithm, the LM algorithm is applied to the command automation networkintrusion classification.Thirdly, to obtain the optimal solution with the help of matlab experiment platform by adjusting the parameters in the algorithm research of the firewall flow prediction andintrusion detection. The thesis compares traditional BP algorithm with wavelet algorithm andLM algorithm through simulation.Finally, to summarize the advantages and disadvantages of different algorithms infirewall flow forecast and intrusion detection, and the problem to be further solved is putforward based on the actual situation faced by the command automation network.