Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

One Framework for All: Cross-Modal Membership Inference for Generative Models

About

Large generative models across text-to-text, text-to-image, and image-to-text modalities have been shown to pose significant privacy risks. One fundamental threat is membership inference attacks (MIA), which aim to determine whether a given data point was used in a model's training set. Although prior work has investigated MIAs against these three classes of generative models, existing approaches treat them in isolation and are not cross-applicable, thereby limiting their real-world utility. To address this limitation, we present the first comprehensive study of a unified membership inference framework that applies across text-to-text, text-to-image, and image-to-text modalities. Our approach is grounded in a key modality-agnostic observation: the output distribution of a generative model can approximate its training data distribution. Leveraging this property, we model the distributions of model-generated outputs and auxiliary non-member samples in a shared embedding space, and perform membership inference via likelihood ratio testing. We conduct extensive experiments in a strict black-box setting under both partial-knowledge and zero-knowledge threat models, and evaluate membership inference against both fine-tuning and pre-training data. Experimental results demonstrate our approach's superior performance in comparison to existing state-of-the-art methods, which are typically optimized for a single model class.

Dayong Ye, Tainqing Zhu, Kun Gao, Junhao Liu, Yichuan Chen, Shuai Zhou, Hengzhu Liu, Bo Liu, Wanlei Zhou• 2026

Related benchmarks

TaskDatasetResultRank
Membership InferenceMS-COCO
ASR86
27
Image-to-Text Membership InferenceCOCO 2017
ASR91
6
Image-to-Text Membership InferenceCC SBU ALIGN
ASR85
6
Text-to-Image Membership InferenceCelebA-Dialog
ASR96
6
Text-to-Text Membership InferenceWiki103
ASR90
6
Text-to-Text Membership InferenceXsum
ASR98
6
Membership InferenceGPT-2 (train)
ASR86
2
Membership InferenceGuided-diffusion Pre-training Data
Attack Success Rate (ASR)96
2
Membership InferenceLLaVA Pre-training
ASR85
2
Showing 9 of 9 rows

Other info

Follow for update