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

The emergence of vertical federated learning (VFL) has stimulated concernsabout the imperfection in privacy protection, as shared feature embeddings mayreveal sensitive information under privacy attacks. This paper studies thedelicate equilibrium between data privacy and task utility goals of VFL underdifferential privacy (DP). To address the generality issue of prior arts, thispaper advocates a flexible and generic approach that decouples the two goalsand addresses them successively. Specifically, we initially derive a rigorousprivacy guarantee by applying norm clipping on shared feature embeddings, whichis applicable across various datasets and models. Subsequently, we demonstratethat task utility can be optimized via adaptive adjustments on the scale anddistribution of feature embeddings in an accuracy-appreciative way, withoutcompromising established DP mechanisms. We concretize our observation into theproposed VFL-AFE framework, which exhibits effectiveness against privacyattacks and the capacity to retain favorable task utility, as substantiated byextensive experiments.

Flexible Differentially Private Vertical Federated Learning with Adaptive Feature Embeddings