Optimizing BiLSTM Network Attack Prediction Based on Improved Gray Wolf Algorithm

Aiming at the problems of low accuracy network attack prediction and long response time detection, bidirectional short-term memory (BiLSTM) was used to predict attacks. However, BiLSTM has difficulty in parameter setting model. This paper first proposes Improved Grey Wolf algorithm (IGWO) optimize (IGWO-BiLSTM). First, IGWO uses Dimension Learning Hunting (DLH) strategy construct wolf neighborhood. In established neighborhood, parameters are iteratively optimized obtain a model with fast convergence speed small reconstruction error. Secondly, dataset is preprocessed, IP packet statistical signature (IPDCF) defined according characteristics denial service (DOS) distributed (DDOS) IPDCF establish series traffic data were input into IGWO-BiLSTM get results. Finally, DOS DDOS packets trained results data. By comparing predicted values normal packets, reasonable threshold set provide basis for subsequent prediction. Experiments show that can reach 99.05% fitting degree accurately distinguish attacks from demand increases.