Abstract
In recent years, with the rise of the global network, the Internet technology as the core of large enterprise network system is developing rapidly. It is widely used in the field of electronic commerce, information service, network communication and other technical means. At the same time, the problem of network security has become increasingly prominent. Due to historical and technical reasons, the security system of enterprise network is still very weak in China. The network basic software and hardware is still use a large number of foreign products, and the core of the network security technology cannot be fully mastered. So, the security risks are obvious. Therefore, strengthening the construction of enterprise network security system and researching the safe application model has become the top priority of our country's enterprise information technology. Based on this, we propose a smoothing method to improve the initial weights and the initial threshold, and use a test method to select the hidden layer node number of neural network. So, we can minimize the fitting error of the training. In this paper, the computer network security data of an enterprise is selected, and all the indexes are scored by the experts. The result of the scoring is the input value of the improved BP neural network. Finally, we use this algorithm to predict the network security of a certain enterprise in the next three months. The score is 0.85, 0.88 and 0.91, which is close to the actual value of network security.
