# When Visual Privacy Protection Meets Multimodal Large Language Models

arXiv: https://arxiv.org/abs/2603.13978 · HTML: https://mlsanity.com/paper/2603.13978/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 21 of its 49 references that are in the map.
- First posted on arXiv 14 Mar 2026.

## Placed in

- Multimodal LLMs / vision-language models / [MLLM jailbreaks and adversarial robustness](https://mlsanity.com/mllm/mllm-jailbreaks-and-adversarial-robustness/)

## Builds on (most cited first, 21 of 21)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [MMBench: Is Your Multi-modal Model an All-around Player?](https://mlsanity.com/paper/2307.06281/) (2023)
- [Video-LLaVA: Learning United Visual Representation by Alignment Before Projection](https://mlsanity.com/paper/2311.10122/) (2023)
- [OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge](https://mlsanity.com/paper/1906.00067/) (2019)
- [Video-ChatGPT: Towards Detailed Video Understanding via Large Vision and Language Models](https://mlsanity.com/paper/2306.05424/) (2023)
- [A Survey on Multimodal Large Language Models](https://mlsanity.com/paper/2306.13549/) (2023)
- [MiniGPT-v2: large language model as a unified interface for vision-language multi-task learning](https://mlsanity.com/paper/2310.09478/) (2023)
- [Delving into Transferable Adversarial Examples and Black-box Attacks](https://mlsanity.com/paper/1611.02770/) (2016)
- [Deep Learning on a Data Diet: Finding Important Examples Early in Training](https://mlsanity.com/paper/2107.07075/) (2021)
- [Scalable Extraction of Training Data from (Production) Language Models](https://mlsanity.com/paper/2311.17035/) (2023)
- [Towards a Visual Privacy Advisor: Understanding and Predicting Privacy Risks in Images](https://mlsanity.com/paper/1703.10660/) (2017)
- [BlackVIP: Black-Box Visual Prompting for Robust Transfer Learning](https://mlsanity.com/paper/2303.14773/) (2023)
- [Unveiling Security, Privacy, and Ethical Concerns of ChatGPT](https://mlsanity.com/paper/2307.14192/) (2023)
- [SPAct: Self-supervised Privacy Preservation for Action Recognition](https://mlsanity.com/paper/2203.15205/) (2022)
- [A Survey on Evaluation of Multimodal Large Language Models](https://mlsanity.com/paper/2408.15769/) (2024)
- [TeD-SPAD: Temporal Distinctiveness for Self-supervised Privacy-preservation for video Anomaly Detection](https://mlsanity.com/paper/2308.11072/) (2023)
- [Selective, Interpretable, and Motion Consistent Privacy Attribute Obfuscation for Action Recognition](https://mlsanity.com/paper/2403.12710/) (2024)
- [Memory Augmented Optimizers for Deep Learning](https://mlsanity.com/paper/2106.10708/) (2021)

## Abstract (arXiv)

The emergence of Multimodal Large Language Models (MLLMs) and the widespread usage of MLLM cloud services such as GPT-4V raised great concerns about privacy leakage in visual data. As these models are typically deployed in cloud services, users are required to submit their images and videos, posing serious privacy risks. However, how to tackle such privacy concerns is an under-explored problem. Thus, in this paper, we aim to conduct a new investigation to protect visual privacy when enjoying the convenience brought by MLLM services. We address the practical case where the MLLM is a "black box", i.e., we only have access to its input and output without knowing its internal model information. To tackle such a challenging yet demanding problem, we propose a novel framework, in which we carefully design the learning objective with Pareto optimality to seek a better trade-off between visual privacy and MLLM's performance, and propose critical-history enhanced optimization to effectively optimize the framework with the black-box MLLM. Our experiments show that our method is effective on different benchmarks.