A Self-Supervised Learning Model for Unknown Internet Traffic Identification Based on Surge Period

The identification of Internet protocols provides a significant basis for keeping security and improving Quality Service (QoS). However, the overwhelming developments updating technologies have led to large volumes unknown traffic, which threaten safety network environment lot. Since most traffic does not any labels, it is difficult adopt deep learning directly. Additionally, feature accuracy model also impact In this paper, we propose surge period-based extraction method that helps remove negative influence background in sessions acquire as many flow features possible. addition, establish an based on JigClu, self-supervised approach training unlabeled datasets. It finally combines with clustering realizes further traffic. has been demonstrated no less than 74% identifying public dataset ISCXVPN2016 under different scenarios. work novel solution identification, task We believe great leap significance maintaining environment.