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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Embedding covert streams into a cover channel is a common approach tocircumventing Internet censorship, due to censors' inability to examineencrypted information in otherwise permitted protocols (Skype, HTTPS, etc.).However, recent advances in machine learning (ML) enable detecting a range ofanti-censorship systems by learning distinct statistical patterns hidden intraffic flows. Therefore, designing obfuscation solutions able to generatetraffic that is statistically similar to innocuous network activity, in orderto deceive ML-based classifiers at line speed, is difficult.  In this paper, we formulate a practical adversarial attack strategy againstflow classifiers as a method for circumventing censorship. Specifically, wecast the problem of finding adversarial flows that will be misclassified as asequence generation task, which we solve with Amoeba, a novel reinforcementlearning algorithm that we design. Amoeba works by interacting with censoringclassifiers without any knowledge of their model structure, but by craftingpackets and observing the classifiers' decisions, in order to guide thesequence generation process. Our experiments using data collected from twopopular anti-censorship systems demonstrate that Amoeba can effectively shapeadversarial flows that have on average 94% attack success rate against a rangeof ML algorithms. In addition, we show that these adversarial flows are robustin different network environments and possess transferability across various MLmodels, meaning that once trained against one, our agent can subvert othercensoring classifiers without retraining.

Amoeba: Circumventing ML-supported Network Censorship via Adversarial Reinforcement Learning