Feature identification and processing strategies of machine learning techniques in big data traffic analysis

In the era of booming Internet, the network affects every aspect of people unconsciously, while illegal attacks in network traffic bring security risks, and traditional traffic detection faces serious challenges. In this study, the stacked self-coding neural network is combined with the twin neural network model to construct the SAE-SCNN model, and the traffic features are extracted by using the convolutional layer and pooling layer in the model. The traffic features are classified according to the calculation results of the distance function, and then piggyback on the method to build a big data traffic analysis model. The performance evaluation test results show that in the public datasets CICIDS2017 and UNSW-NB15, the detection performance of this paper for traffic shows significant improvement compared to other models. For normal data traffic in the simulation test, the model detection accuracy can reach up to 99.99%, and it also has high accuracy and generalization for abnormal traffic detection.