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

Membership Inference Attacks against Large Vision-Language Models

About

Large vision-language models (VLLMs) exhibit promising capabilities for processing multi-modal tasks across various application scenarios. However, their emergence also raises significant data security concerns, given the potential inclusion of sensitive information, such as private photos and medical records, in their training datasets. Detecting inappropriately used data in VLLMs remains a critical and unresolved issue, mainly due to the lack of standardized datasets and suitable methodologies. In this study, we introduce the first membership inference attack (MIA) benchmark tailored for various VLLMs to facilitate training data detection. Then, we propose a novel MIA pipeline specifically designed for token-level image detection. Lastly, we present a new metric called MaxR\'enyi-K%, which is based on the confidence of the model output and applies to both text and image data. We believe that our work can deepen the understanding and methodology of MIAs in the context of VLLMs. Our code and datasets are available at https://github.com/LIONS-EPFL/VL-MIA.

Zhan Li, Yongtao Wu, Yihang Chen, Francesco Tonin, Elias Abad Rocamora, Volkan Cevher• 2024

Related benchmarks

TaskDatasetResultRank
Image Membership Inference AttackVL-MIA Flickr (test)
AUC0.937
149
Membership Inference AttackFlickr
TPR @ 5% FPR23
142
Membership InferenceVL-MIA DALL·E
AUROC86.5
120
Text Membership Inference AttackLLaVA LLM Pre-training
AUC0.688
88
Membership Inference AttackFlickr
Accuracy67.5
71
Membership Inference AttackDALL·E (test)
TPR @ 5% FPR22.3
54
Membership Inference AttackVL-MIA Flickr 2k
AUC0.668
45
Membership Inference AttackVL-MIA Flickr-10k
AUC0.676
45
Membership Inference AttackVL-MIA Flickr v1 (test)
AUC0.627
45
Text Membership Inference AttackLLaVA VLLM Tuning
AUC0.993
44
Showing 10 of 29 rows

Other info

Code

Follow for update