A method of intrusion detection based on Attention-LSTM neural network
Shuaichuang Yang, Minsheng Tan, Shiying Xia, Fangju Liu
Abstract
Recently, network attacks with complex types have occurred more frequently than before, and traditional detection algorithms cannot meet current needs. For this reason, an intrusion detection method based on Attention- Long Short Term Memory (LSTM) neural network is proposed. This method combines the advantage of the attention mechanism theory to solve the problem of the inability to pay attention to key attributes in intrusion detection. At the same time, it uses the memory function of the Long Short Term Memory network and powerful series data learning ability to learn. Finally, the KDD-CUP99 data sets are used to test the performance of attention-LSTM. The experiment results show that the proposed algorithm is efficient. Compare with the classical Convolutional Neural Networks (CNN) algorithm, Recurrent Neural Network (RNN) algorithm, and LSTM algorithm, the method not only improves the accuracy rate and precision rate of network intrusion detection but also decreases the false alarm rate. It provides a design basis and technical support for future intrusion detection technology.