Salvato in:
Dettagli Bibliografici
Autori principali: Sun, Xiongtao, Li, Hui, Zhang, Jiaming, Yang, Yujie, Liu, Kaili, Feng, Ruxin, Tan, Wen Jun, Lim, Wei Yang Bryan
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2511.16940
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914615097556992
author Sun, Xiongtao
Li, Hui
Zhang, Jiaming
Yang, Yujie
Liu, Kaili
Feng, Ruxin
Tan, Wen Jun
Lim, Wei Yang Bryan
author_facet Sun, Xiongtao
Li, Hui
Zhang, Jiaming
Yang, Yujie
Liu, Kaili
Feng, Ruxin
Tan, Wen Jun
Lim, Wei Yang Bryan
contents Modern Vision-Language Models (VLMs) pose significant individual-level privacy risks by linking fragmented multimodal data to identifiable individuals through hierarchical chain-of-thought reasoning. However, existing privacy benchmarks remain structurally insufficient for this threat, as they primarily evaluate privacy perception while failing to address the more critical risk of privacy reasoning: a VLM's ability to infer and link distributed information to construct individual profiles. To address this gap, we propose MultiPriv, the first benchmark designed to systematically evaluate individual-level privacy reasoning in VLMs. We introduce the Privacy Perception and Reasoning (PPR) framework and construct a bilingual multimodal dataset with synthetic individual profiles, where identifiers, such as faces and names, are linked to sensitive attributes. This design enables nine challenging tasks spanning attribute detection, cross-image re-identification, and chained inference. We conduct a large-scale evaluation of over 50 open-source and commercial VLMs. In our controlled benchmark, 60% of widely used VLMs can perform individual-level privacy reasoning with up to 80% accuracy, suggesting a significant potential threat to personal privacy. The benchmark is available at https://github.com/CyberChangAn/MultiPriv-PII.
format Preprint
id arxiv_https___arxiv_org_abs_2511_16940
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MultiPriv: Benchmarking Individual-Level Privacy Reasoning in Vision-Language Models
Sun, Xiongtao
Li, Hui
Zhang, Jiaming
Yang, Yujie
Liu, Kaili
Feng, Ruxin
Tan, Wen Jun
Lim, Wei Yang Bryan
Computer Vision and Pattern Recognition
Cryptography and Security
Modern Vision-Language Models (VLMs) pose significant individual-level privacy risks by linking fragmented multimodal data to identifiable individuals through hierarchical chain-of-thought reasoning. However, existing privacy benchmarks remain structurally insufficient for this threat, as they primarily evaluate privacy perception while failing to address the more critical risk of privacy reasoning: a VLM's ability to infer and link distributed information to construct individual profiles. To address this gap, we propose MultiPriv, the first benchmark designed to systematically evaluate individual-level privacy reasoning in VLMs. We introduce the Privacy Perception and Reasoning (PPR) framework and construct a bilingual multimodal dataset with synthetic individual profiles, where identifiers, such as faces and names, are linked to sensitive attributes. This design enables nine challenging tasks spanning attribute detection, cross-image re-identification, and chained inference. We conduct a large-scale evaluation of over 50 open-source and commercial VLMs. In our controlled benchmark, 60% of widely used VLMs can perform individual-level privacy reasoning with up to 80% accuracy, suggesting a significant potential threat to personal privacy. The benchmark is available at https://github.com/CyberChangAn/MultiPriv-PII.
title MultiPriv: Benchmarking Individual-Level Privacy Reasoning in Vision-Language Models
topic Computer Vision and Pattern Recognition
Cryptography and Security
url https://arxiv.org/abs/2511.16940