When Privacy Meets Recovery: The Overlooked Half of Surrogate-Driven Privacy Preservation for MLLM Editing
It addresses the overlooked issue of privacy recovery quality in MLLM editing, which is crucial for real-world applications where user data authenticity matters.
This work tackles the problem of evaluating and restoring private information in Multimodal Large Language Models (MLLMs) after it has been obscured for privacy, introducing a dataset and method that achieve a strong balance between privacy protection and MLLM usability across diverse scenarios.
Privacy leakage in Multimodal Large Language Models (MLLMs) has long been an intractable problem. Existing studies, though effectively obscure private information in MLLMs, often overlook the evaluation of the authenticity and recovery quality of user privacy. To this end, this work uniquely focuses on the critical challenge of how to restore surrogate-driven protected data in diverse MLLM scenarios. We first bridge this research gap by contributing the SPPE (Surrogate Privacy Protected Editable) dataset, which includes a wide range of privacy categories and user instructions to simulate real MLLM applications. This dataset offers protected surrogates alongside their various MLLM-edited versions, thus enabling the direct assessment of privacy recovery quality. By formulating privacy recovery as a guided generation task conditioned on complementary multimodal signals, we further introduce a unified approach that reliably reconstructs private content while preserving the fidelity of MLLM-generated edits. The experiments on both SPPE and InstructPix2Pix further show that our approach generalizes well across diverse visual content and editing tasks, achieving a strong balance between privacy protection and MLLM usability.