Article Preview
Top1. Introduction
Physical fusion of information and deep learning are rapidly developing technologies in recent years, and they have a wide application prospect in network communication information security. Information physical fusion can combine physical signal with information processing technology to improve the reliability and security of data transmission. Deep learning has powerful data processing and analysis capabilities, which can monitor and predict the data in network communication in real time and improve the security of the network. Data in network communication is usually sequential data that arrives in chronological order, such as network traffic and log records. LSTM can effectively capture long-term dependencies in sequence, model and predict sequence data. In the field of information security, LSTM can be used to detect abnormal network behavior, intrusion detection and password cracking. CNN is widely used in the field of image processing, but its excellent feature extraction ability can also be used in network communication information security. Through convolutional operation, CNN can extract local and spatial features in data, which are often important information in network communication data, such as packet header information in network traffic and characters on the login page. CNN can also be combined with other network level information for classification and identification, such as the use of convolutional layer and full connection layer for spam filtering, malicious URL detection and other tasks.
With continuous research on fifth-generation mobile communication networks, the number of various terminals and the demand for services are exploding, and the operation of wireless networks is facing many challenges. The traditional manual management method is no longer suitable for this network structure. Therefore, timely detection and diagnosis of network faults through intelligent algorithms under the premise of ensuring system capacity has become one of the focuses of current network operators. Given that the improvement of user experience is very important for network service providers, effective solutions are needed in the field of network security diagnostics, which would surely be a research topic for future network security maintenance. The innovation of this paper is that it proposes a deep learning method for cyber-physical fusion and elaborates and compares the LSTM and CNN methods in detail. The use of a reputation mechanism to protect network security is proposed.
Although many scholars have conducted in-depth research on LSTM and CNN methods and there are many application areas for network security assurance, the variation patterns of network fault prediction accuracy under different quantities have not been explored. However, this article found through experiments that as the number of data records increases, the accuracy on the training set gradually improves, the stability improves, and the accuracy on the test set significantly improves. However, when the historical data were recorded at 400, there was almost no improvement in performance.