BiLSTM Model Based on Bayesian Optimization for Network Security Situation Prediction

With the continuous expansion of network scale and the gradual deepening of complexity, network attack techniques are also in the process of dynamic evolution. As a core hot spot in the field of network security, network security situation prediction is receiving more and more widespread attention. Traditional network security situation prediction methods have many limitations, such as single approach, poor prediction accuracy, insufficient generalization ability, and high time cost. In this paper, we take the bidirectional long and short-term memory network (BiLSTM) as the main research object, optimize the BiLSTM with the help of Bayesian optimization algorithm (BO) and attention mechanism. Then propose and construct a new forecasting model BO - BiLSTM - ATT (BBLA), and carry out the relevant experimental research. Through simulation experiments on the UNSW-NB15 network traffic dataset, the predicted situation values of the optimized BiLSTM are highly consistent with the actual posture values, which fully verifies the effectiveness and superiority of the proposed method.