Research on network security situation prediction method based on AM and LSTM hybrid neural network

Network security situation prediction is an important part of network security situation awareness. Traditional network security situation prediction methods are insensitive to the characteristics of input data, and there are also important degrees of variability and temporal correlation among data. To address these problems, a network security situation prediction model that combines attention mechanism (AM) with convolutional neural network (CNN) and long short-term memory (LSTM) is proposed. Using time-series data as input, CNN is used to extract the input features of the data to obtain a high-level feature representation; LSTM is used to capture the short-term mutations of time series for prediction. AM is introduced to assign different weights to the input features before LSTM prediction to improve the model accuracy and find more useful information for prediction. Through experimental comparison, the method has better prediction accuracy compared with the three models of CNN-LSTM, AM-LSTM and LSTM, indicating that the proposed method is more suitable for application in network security situation prediction and has great potential in network security situation prediction.