A Network Intrusion Detection Algorithm Based on Adversarial Dual Channel Encoder
Aiming at the low detection rate of minority attacks caused by the imbalance of network traffic data,an intrusion detection algorithm based on adversarial dual channel encoder is proposed.Normal and attack traffic are used respectively to train the variational autoencoder model to construct a new feature vector based on a multi-channel representation of the autoencoder-derived traffic data.The generative process of generating the adversarial network is driven to develop towards the target class,a small number of class images is generated,the dataset is effectively extended.The feature extraction capability of the model is enhanced by introducing a CBAM module to improve the network structure of the generator,fusing features in both channel and spatial directions.The discriminator output is adjusted to a single target classification and a softmax layer is added to output Fake,Normal and Attack results to avoid the generator generating images that cannot be rewarded for matching the desired type and to improve the quality of the generated images.The experimental results show that the proposed method can effectively reduce the false alarm rate and can improve the detection accuracy of unknown attacks,can have more superiorities especially in unbalanced data sets.
intrusion detection algorithmauxiliary generation of adversarial networkautoencoderattention mechanism
国家科技期刊平台
NSTL
万方数据
