Identity-Decoupled Anonymization for Visual Evidence in Multi-modal Retrieval-Augmented Generation

arXiv 2604.23584 ↗First posted 26 Apr 2026cs.CV cs.IRcited by 0

On the map: Multimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness

also in 2 more

Abstract (arXiv)

Multi-modal retrieval-augmented generation (MRAG) systems retrieve visual evidence from large image corpora to ground the responses of large multi-modal models, yet the retrieved images frequently contain human faces whose identities constitute sensitive personal information. Existing anonymization techniques that destroy the non-identity visual cues that downstream reasoning depends on or fail to provide principled privacy guarantees. We propose Identity-Decoupled MRAG, a framework that interposes a generative anonymization module between retrieval and generation. Our approach consists of three components: (i)a disentangled variational encoder that factorizes each face into an identity code and a spatially-structured attribute code, regularized by a mutual-information penalty and a gradient-based independence term; (ii)a manifold-aware rejection sampler that replaces the identity code with a synthetic one guaranteed to be both distinct from the original and realistic; and (iii)a conditional latent diffusion generator that synthesizes the anonymized face from the replacement identity and the preserved attributes, distilled into a latent consistency model for low-latency deployment. Privacy is enforced through a multi-oracle ensemble of face recognition models with a hinge-based loss that halts optimization once identity similarity drops below the impostor-regime threshold.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 17 of its 42 references that are in the map. First posted on arXiv 26 Apr 2026.

Builds on

17 of 42 references.

References · 17

2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 5,986
2016SGDR: Stochastic Gradient Descent with Warm Restartscited by 4,993
2014Deep Learning Face Attributes in the Wildcited by 3,679
2018Spectral Normalization for Generative Adversarial Networkscited by 1,618
2018CosFace: Large Margin Cosine Loss for Deep Face Recognitioncited by 968
2016Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networkscited by 895
2018MINE: Mutual Information Neural Estimationcited by 680
2018Disentangling by Factorisingcited by 614
2022AdaFace: Quality Adaptive Margin for Face Recognitioncited by 390
2017Variational Inference of Disentangled Latent Concepts from Unlabeled Observationscited by 240
2020Disentangled and Controllable Face Image Generation via 3D Imitative-Contrastive Learningcited by 219

5 more

2022MuRAG: Multimodal Retrieval-Augmented Generator for Open Question Answering over Images and Textcited by 216
2019DeepPrivacy: A Generative Adversarial Network for Face Anonymizationcited by 108
2020CIAGAN: Conditional Identity Anonymization Generative Adversarial Networkscited by 88
2022L2CS-Net: Fine-Grained Gaze Estimation in Unconstrained Environmentscited by 65
2025Beyond Inference Intervention: Identity-Decoupled Diffusion for Face Anonymizationcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way