To address the low accuracy in encrypted traffic recognition and inefficiency in model per-formance,a novel one-dimensional convolutional neural network(1D-CNN)model integrated with Transformer technology is proposed.Firstly,the superior performance of convolutional neural networks in sequence data analysis is combined with the powerful representational capabilities of Transformers for sequence relationships.Local features in encrypted traffic data are deeply mined by employing two one-dimensional convolutional neural networks.Secondly,Transformer neural network embedding and position encoding strategies are simultaneously used,and complex data features are transformed into semantic vectors.Thirdly,multi-head attention mechanisms are utilized to further enhance its capabil-ity to capture deep inter-sequence dependencies,and global features are efficiently extracted from en-crypted data.Finally,the fusion of these features occurs in a fully connected layer,followed by the use of a classifier for the recognition of encrypted traffic attributes.It is validated on the ISCX 2016 VPN-nonVPN dataset.Experimental results demonstrate that this model achieves an accuracy of 96.7%,in-dicating a significant performance improvement in the task of encrypted traffic classification.
deep learningencrypted traffic classificationone-dimensional convolutional neural net-workmulti-head attention mechanismintegrated model
维普
万方数据
