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

In Member Inference (MI) attacks, the adversary try to determine whether aninstance is used to train a machine learning (ML) model. MI attacks are a majorprivacy concern when using private data to train ML models. Most MI attacks inthe literature take advantage of the fact that ML models are trained to fit thetraining data well, and thus have very low loss on training instances. Mostdefenses against MI attacks therefore try to make the model fit the trainingdata less well. Doing so, however, generally results in lower accuracy. Weobserve that training instances have different degrees of vulnerability to MIattacks. Most instances will have low loss even when not included in training.For these instances, the model can fit them well without concerns of MIattacks. An effective defense only needs to (possibly implicitly) identifyinstances that are vulnerable to MI attacks and avoids overfitting them. Amajor challenge is how to achieve such an effect in an efficient trainingprocess. Leveraging two distinct recent advancements in representationlearning: counterfactually-invariant representations and subspace learningmethods, we introduce a novel Membership-Invariant Subspace Training (MIST)method to defend against MI attacks. MIST avoids overfitting the vulnerableinstances without significant impact on other instances. We have conductedextensive experimental studies, comparing MIST with various otherstate-of-the-art (SOTA) MI defenses against several SOTA MI attacks. We findthat MIST outperforms other defenses while resulting in minimal reduction intesting accuracy.

MIST: Defending Against Membership Inference Attacks Through Membership-Invariant Subspace Training