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

As in-the-wild data are increasingly involved in the training stage, machinelearning applications become more susceptible to data poisoning attacks. Suchattacks typically lead to test-time accuracy degradation or controlledmisprediction. In this paper, we investigate the third type of exploitation ofdata poisoning - increasing the risks of privacy leakage of benign trainingsamples. To this end, we demonstrate a set of data poisoning attacks to amplifythe membership exposure of the targeted class. We first propose a genericdirty-label attack for supervised classification algorithms. We then propose anoptimization-based clean-label attack in the transfer learning scenario,whereby the poisoning samples are correctly labeled and look "natural" to evadehuman moderation. We extensively evaluate our attacks on computer visionbenchmarks. Our results show that the proposed attacks can substantiallyincrease the membership inference precision with minimum overall test-timemodel performance degradation. To mitigate the potential negative impacts ofour attacks, we also investigate feasible countermeasures.

Amplifying Membership Exposure via Data Poisoning