Assessing Privacy Preservation and Utility in Online Vision-Language Models

TL;DR

This paper examines privacy risks in online vision-language models, focusing on PII exposure through images, and proposes methods to balance privacy protection with utility preservation.

Contribution

It identifies privacy vulnerabilities in OVLMs related to PII disclosure and introduces techniques to safeguard privacy without compromising image utility.

Findings

Proposed privacy-preserving methods effectively reduce PII exposure.

Evaluation shows a balance between privacy protection and utility retention.

Highlighting the importance of contextual relationship analysis in privacy risks.

Abstract

The increasing use of Online Vision Language Models (OVLMs) for processing images has introduced significant privacy risks, as individuals frequently upload images for various utilities, unaware of the potential for privacy violations. Images contain relationships that relate to Personally Identifiable Information (PII), where even seemingly harmless details can indirectly reveal sensitive information through surrounding clues. This paper explores the critical issue of PII disclosure in images uploaded to OVLMs and its implications for user privacy. We investigate how the extraction of contextual relationships from images can lead to direct (explicit) or indirect (implicit) exposure of PII, significantly compromising personal privacy. Furthermore, we propose methods to protect privacy while preserving the intended utility of the images in Vision Language Model (VLM)-based applications. Our evaluation demonstrates the efficacy of these techniques, highlighting the delicate balance between maintaining utility and protecting privacy in online image processing environments. Index Terms-Personally Identifiable Information (PII), Privacy, Utility, privacy concerns, sensitive information