Build IPSO-ABiLSTM Model for Network Security Situation Prediction

There are security risks in interaction and communication using wireless mobile networks, and network security situation prediction technology is to predict the next development trend with the previous and current network status, which can grasp the wireless mobile networks security status in time and make decisions in advance to avoid attacks. This paper proposes an Improved Particle Swarm Optimization Attention Bidirectional Long Short-Term Memory (IPSO-ABiLSTM) model for network security situation prediction. First, we construct the real situation values of the raw UNSW-NB15 dataset from the perspective of the impact of the attack on the situation indicator system, the sliding window method was introduced to reconstruct the situation values of the data set obtained by computing the data used for prediction. Secondly, the traditional PSO algorithm has the shortage of unbalanced search speed and tends to get local optimal solutions. The IPSO algorithm in this paper makes the global and local search ability of the algorithm more balanced and converges faster. Finally, the IPSO-ABiLSTM model is used to implement the situation prediction in different sliding window sizes. The experimental results show that the IPSO-ABiLSTM of this paper fits up to 0.9922, which verifies the effectiveness of the model proposed in this paper in the network situation prediction problem.