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

Membership Inference Attacks (MIA) aim to infer whether a target data recordhas been utilized for model training or not. Prior attempts have quantified theprivacy risks of language models (LMs) via MIAs, but there is still noconsensus on whether existing MIA algorithms can cause remarkable privacyleakage on practical Large Language Models (LLMs). Existing MIAs designed forLMs can be classified into two categories: reference-free and reference-basedattacks. They are both based on the hypothesis that training recordsconsistently strike a higher probability of being sampled. Nevertheless, thishypothesis heavily relies on the overfitting of target models, which will bemitigated by multiple regularization methods and the generalization of LLMs.The reference-based attack seems to achieve promising effectiveness in LLMs,which measures a more reliable membership signal by comparing the probabilitydiscrepancy between the target model and the reference model. However, theperformance of reference-based attack is highly dependent on a referencedataset that closely resembles the training dataset, which is usuallyinaccessible in the practical scenario. Overall, existing MIAs are unable toeffectively unveil privacy leakage over practical fine-tuned LLMs that areoverfitting-free and private. We propose a Membership Inference Attack based onSelf-calibrated Probabilistic Variation (SPV-MIA). Specifically, sincememorization in LLMs is inevitable during the training process and occursbefore overfitting, we introduce a more reliable membership signal,probabilistic variation, which is based on memorization rather thanoverfitting. Furthermore, we introduce a self-prompt approach, which constructsthe dataset to fine-tune the reference model by prompting the target LLMitself. In this manner, the adversary can collect a dataset with a similardistribution from public APIs.

Practical Membership Inference Attacks against Fine-tuned Large Language Models via Self-prompt Calibration