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

While being deployed in many critical applications as core components,machine learning (ML) models are vulnerable to various security and privacyattacks. One major privacy attack in this domain is membership inference, wherean adversary aims to determine whether a target data sample is part of thetraining set of a target ML model. So far, most of the current membershipinference attacks are evaluated against ML models trained from scratch.However, real-world ML models are typically trained following the transferlearning paradigm, where a model owner takes a pretrained model learned from adifferent dataset, namely teacher model, and trains her own student model byfine-tuning the teacher model with her own data.  In this paper, we perform the first systematic evaluation of membershipinference attacks against transfer learning models. We adopt the strategy ofshadow model training to derive the data for training our membership inferenceclassifier. Extensive experiments on four real-world image datasets show thatmembership inference can achieve effective performance. For instance, on theCIFAR100 classifier transferred from ResNet20 (pretrained with Caltech101), ourmembership inference achieves $95\%$ attack AUC. Moreover, we show thatmembership inference is still effective when the architecture of target modelis unknown. Our results shed light on the severity of membership risks stemmingfrom machine learning models in practice.

Privacy Analysis of Deep Learning in the Wild: Membership Inference Attacks against Transfer Learning