SecureT2I: No More Unauthorized Manipulation on AI Generated Images from Prompts

2025-07-04

AI Summary

This paper addresses the misuse of text-guided image editing in diffusion models, which poses significant copyright and ethical risks. To mitigate these concerns, we propose the first security framework for edit permission control. Our core innovation lies in partitioning image regions into “editable” and “non-editable” sets, enforced via a dual-path loss function: the non-editable set is constrained to produce semantically ambiguous outputs—achieved through lightweight blurring strategies (e.g., resize-based degradation)—while the editable set retains high-fidelity, prompt-aligned editing. The method requires only light fine-tuning and is compatible with mainstream prompt-driven diffusion models. Experiments across multiple datasets and architectures demonstrate that our approach substantially degrades unauthorized editing quality (average reduction of 42.7% in perceptual fidelity), with negligible impact on legitimate editing performance (PSNR drop < 0.3 dB). Moreover, it exhibits strong generalization across unseen models and editing scenarios.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Safety and Robustness
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Measurement, analysis, and circumvention of Web censorship

Abstract

Text-guided image manipulation with diffusion models enables flexible and precise editing based on prompts, but raises ethical and copyright concerns due to potential unauthorized modifications. To address this, we propose SecureT2I, a secure framework designed to prevent unauthorized editing in diffusion-based generative models. SecureT2I is compatible with both general-purpose and domain-specific models and can be integrated via lightweight fine-tuning without architectural changes. We categorize images into a permit set and a forbid set based on editing permissions. For the permit set, the model learns to perform high-quality manipulations as usual. For the forbid set, we introduce training objectives that encourage vague or semantically ambiguous outputs (e.g., blurred images), thereby suppressing meaningful edits. The core challenge is to block unauthorized editing while preserving editing quality for permitted inputs. To this end, we design separate loss functions that guide selective editing behavior. Extensive experiments across multiple datasets and models show that SecureT2I effectively degrades manipulation quality on forbidden images while maintaining performance on permitted ones. We also evaluate generalization to unseen inputs and find that SecureT2I consistently outperforms baselines. Additionally, we analyze different vagueness strategies and find that resize-based degradation offers the best trade-off for secure manipulation control.

Problem

• Prevent unauthorized editing in diffusion-based image generation models
• Maintain high-quality edits for permitted images while blocking forbidden ones
• Achieve secure manipulation control via selective training objectives

Innovation

• Lightweight fine-tuning for secure diffusion models
• Separate loss functions for selective editing
• Resize-based degradation for secure manipulation control