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

A large body of research has shown that machine learning models arevulnerable to membership inference (MI) attacks that violate the privacy of theparticipants in the training data. Most MI research focuses on the case of asingle standalone model, while production machine-learning platforms oftenupdate models over time, on data that often shifts in distribution, giving theattacker more information. This paper proposes new attacks that take advantageof one or more model updates to improve MI. A key part of our approach is toleverage rich information from standalone MI attacks mounted separately againstthe original and updated models, and to combine this information in specificways to improve attack effectiveness. We propose a set of combination functionsand tuning methods for each, and present both analytical and quantitativejustification for various options. Our results on four public datasets showthat our attacks are effective at using update information to give theadversary a significant advantage over attacks on standalone models, but alsocompared to a prior MI attack that takes advantage of model updates in arelated machine-unlearning setting. We perform the first measurements of theimpact of distribution shift on MI attacks with model updates, and show that amore drastic distribution shift results in significantly higher MI risk than agradual shift. Our code is available athttps://www.github.com/stanleykywu/model-updates.

How to Combine Membership-Inference Attacks on Multiple Updated Models