Federated Traffic Synthesizing And Classification Using Generative Adversarial Networks

With the rapid development of wireless communication technology and fog computing,new services and applications are emerging in an endless stream,resulting in more and more complex traffic data categories generated at the edge of the network.Considering that network operators need to meet different requirements for bandwidth and delay in the network according to different traffic,traffic classification and identification are considered to be the key research directions for accurately assessing network performance and effectively managing network resources.This paper first reviews some common methods in the field of traffic classification.With the increasing demand for emerging services and applications and the increasing awareness of data confidentiality,traditional centralized traffic classification methods are facing unprecedented challenges.The rapid growth of data in wireless networks has further challenged traditional cloud computing-based processing methods,the traditional method based on cloud computing requires a large number of labeled data sets,which not only means high costs,but may also lead to information leakage.In order to meet the requirements of low latency and high reliability of services,fog computing is considered to be a good supplement to cloud computing,the rise of fog computing has allowed more data to be distributed and stored at the edge of the network,this paper conducts an in-depth investigation of traffic classification under the distributed architecture of fog computing.Although the existing federated learning methods can identify distributed datasets under the premise of ensuring data security,most of them are based on supervised learning.In this way,it is still difficult to classify distributed unlabeled datasets.Aiming at the pain points of classifying distributed unlabeled datasets in traffic classification,this paper introduces a novel framework,Federated Generative Adversarial Networks and Automatic Classification(FGAN-AC),which integrates decentralized data synthsizing with traffic classification.FGAN-AC is able to synthesize and classify multiple types of service data traffic from decentralized local datasets without requiring a large volume of manually labeled dataset or causing any data leakage.Two types of data synthesizing approaches have been proposed and compared: computation-efficient FGAN(FGAN-?)and communication-efficient FGAN(FGAN-?).The former only implements a single convolutional neural network model for processing each local dataset to reduce the computing resources consumed and the later only requires coordination of intermediate model training parameters to achieve the purpose of saving communication resources.An automatic data classification and model updating framework has been proposed to automatically identify unknown traffic from the synthesized data samples and create new pseudo-labels for model training.Numerical results show that our proposed framework has the ability to synthesize highly mixed service data traffic and can significantly improve the traffic classification performance compared to existing solutions.