Font Size: a A A

Research On BiLSTM Network Intrusion Detection Method Based On Improved Attention Mechanism

Posted on:2021-08-01Degree:MasterType:Thesis
Country:ChinaCandidate:X D LiuFull Text:PDF
GTID:2568306632968169Subject:Systems Engineering
Abstract/Summary:PDF Full Text Request
With the rapid development of network technology,people use the network more frequently,and the problem of network security has gradually become an increasingly serious problem.Intrusion detection is a commonly used detection method in network security.In recent years,it has become an important means of defending against network threats.Network intrusion detection systems of various sizes are widely used in enterprises and governments.Most traditional intrusion detection methods are based on rules matching,statistics and other methods,and the detection effect is not good when faced with a large amount of data.In recent years,machine learning and deep learning methods have been applied to the field of intrusion detection and have become a popular research direction.Based on the analysis of commonly used intrusion detection algorithms,feature processing methods and test data sets of intrusion detection methods,this paper proposes a two-way long-term memory network intrusion detection method based on attention mechanism.The main work of this paper is as follows:1.Analyze the KDD CUP 99 data set,and according to the problem of high network data delay requirements and high data dimensions,use feature selection and feature extraction methods for feature processing,and feature extraction and features based on the idea of feature fusion The selected features are fused and further combined with the multi-layer perceptron classification algorithm to verify the effectiveness of the feature fusion method.2.Aiming at the temporal characteristics of network data,a bidirectional long-term and short-term memory network with attention mechanism is applied to network intrusion detection,and an intrusion detection framework based on the AM-BiLSTM network is constructed.Through experimental analysis under different network structures,The model has multiple detection classification indicators such as accuracy,precision,and recall.Experimental results show that the model has better detection effect than the detection algorithm based on long-short-term memory network and convolutional neural network.3.Aiming at the problem that AM-LSTM does not pay enough attention to the characteristics of input sub-windows in multiple time steps,this paper introduces the PSO algorithm and proposes a method to improve attention,giving different attention to input data at different time steps.The force weights are compared with the hierarchical attention model,and experiments show that the method is effective.
Keywords/Search Tags:Network intrusion detection, feature fusion, Bi-directional long-short term memory Network, PSO attention mechanism
PDF Full Text Request
Related items