English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Mobile Internet has profoundly reshaped modern lifestyles in various aspects.Encrypted Traffic Classification (ETC) naturally plays a crucial role inmanaging mobile Internet, especially with the explosive growth of mobile appsusing encrypted communication. Despite some existing learning-based ETC methodsshowing promising results, three-fold limitations still remain in real-worldnetwork environments, 1) label bias caused by traffic class imbalance, 2)traffic homogeneity caused by component sharing, and 3) training with relianceon sufficient labeled traffic. None of the existing ETC methods can address allthese limitations. In this paper, we propose a novel Pre-trAiningSemi-Supervised ETC framework, dubbed PASS. Our key insight is to resample theoriginal train dataset and perform contrastive pre-training without usingindividual app labels directly to avoid label bias issues caused by classimbalance, while obtaining a robust feature representation to differentiateoverlapping homogeneous traffic by pulling positive traffic pairs closer andpushing negative pairs away. Meanwhile, PASS designs a semi-supervisedoptimization strategy based on pseudo-label iteration and dynamic lossweighting algorithms in order to effectively utilize massive unlabeled trafficdata and alleviate manual train dataset annotation workload. PASS outperformsstate-of-the-art ETC methods and generic sampling approaches on four publicdatasets with significant class imbalance and traffic homogeneity, remarkablypushing the F1 of Cross-Platform215 with 1.31%, ISCX-17 with 9.12%.Furthermore, we validate the generality of the contrastive pre-training andpseudo-label iteration components of PASS, which can adaptively benefit ETCmethods with diverse feature extractors.

Listen to Minority: Encrypted Traffic Classification for Class Imbalance with Contrastive Pre-Training