Safe Alone, Unsafe Together: Safeguarding Against Implicit Toxicity When Benign Images Combine

• URL: http://arxiv.org/abs/2607.00576v1
• Date: Wed, 01 Jul 2026 07:59:45 GMT
• Title: Safe Alone, Unsafe Together: Safeguarding Against Implicit Toxicity When Benign Images Combine
• Abstract summary: Multi-image implicit toxicity (MIIT) is a new safety issue where each image appears benign in isolation, but harmful semantics emerge when the images are interpreted jointly.<n>This paper aims to study how to identify MIIT.<n>We first provide a formal definition of MIIT and analyze three key challenges for its detection.<n>We construct MIIT-dataset, an image-only multi-image safety dataset covering seven representative risk categories through an automatic generation pipeline.<n>Finally, we train MiShield with progressively distilled reasoning supervision, enabling it to produce safety judgments accompanied by explicit analyses of the correlated entities that result in the hazards
• Score: 43.604242902855155
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multi-image content has become an increasingly prevalent form of visual communication in social media, giving rise to a new safety issue, multi-image implicit toxicity (MIIT), where each image appears benign in isolation, but harmful semantics emerge when the images are interpreted jointly. MIIT is particularly challenging for existing commercial moderation APIs and models due to the lack of explicit risky cues in each image. This paper aims to study how to identify MIIT. We first provide a formal definition of MIIT and analyze three key challenges for its detection. To alleviate the scarcity of data in this area, we construct MIIT-dataset, an image-only multi-image safety dataset covering seven representative risk categories through an automatic generation pipeline. Finally, we train MiShield with progressively distilled reasoning supervision, enabling it to produce safety judgments accompanied by explicit analyses of the correlated entities that result in the hazards. Experiments show that MiShield-8B models outperform representative moderation services and even larger-scale models, revealing its effectiveness and practical value for this widely used visual format. Warning: This paper contains potentially sensitive content.

Related papers

• Unified Face Attack Detection via Fine-Grained Semantic Guidance [14.035934093163057]
  We build upon the large-scale MS-UFAD dataset which contains over 8 million attack images.<n>We propose a Dual Alignment Forgery Network(DAF-Net) to better leverage these textual information.
  arXiv  Detail & Related papers  (2026-07-09T06:50:31Z)
• The Side Effects of Being Smart: Safety Risks in MLLMs' Multi-Image Reasoning [46.156246746700894]
  We introduce MIR-SafetyBench, the first benchmark focused on multi-image reasoning safety.<n>We find that models with more advanced multi-image reasoning can be more vulnerable on MIR-SafetyBench.<n> unsafe generations exhibit lower attention entropy than safe ones on average.
  arXiv  Detail & Related papers  (2026-01-20T16:24:18Z)
• SafetyPairs: Isolating Safety Critical Image Features with Counterfactual Image Generation [5.313750874857107]
  We introduce SafetyPairs, a framework for generating counterfactual pairs of images that differ only in the features relevant to the given safety policy.<n>Using SafetyPairs, we construct a new safety benchmark, which serves as a powerful source of evaluation data.<n>We release a benchmark containing over 3,020 SafetyPair images spanning a diverse taxonomy of 9 safety categories.
  arXiv  Detail & Related papers  (2025-10-24T03:19:48Z)
• SafeGuider: Robust and Practical Content Safety Control for Text-to-Image Models [74.11062256255387]
  Text-to-image models are highly vulnerable to adversarial prompts, which can bypass safety measures and produce harmful content.<n>We introduce SafeGuider, a two-step framework designed for robust safety control without compromising generation quality.<n>SafeGuider demonstrates exceptional effectiveness in minimizing attack success rates, achieving a maximum rate of only 5.48% across various attack scenarios.
  arXiv  Detail & Related papers  (2025-10-05T10:24:48Z)
• TRCE: Towards Reliable Malicious Concept Erasure in Text-to-Image Diffusion Models [53.937498564603054]
  Recent advances in text-to-image diffusion models enable photorealistic image generation, but they also risk producing malicious content, such as NSFW images.<n>To mitigate risk, concept erasure methods are studied to facilitate the model to unlearn specific concepts.<n>We propose TRCE, using a two-stage concept erasure strategy to achieve an effective trade-off between reliable erasure and knowledge preservation.
  arXiv  Detail & Related papers  (2025-03-10T14:37:53Z)
• Rethinking Bottlenecks in Safety Fine-Tuning of Vision Language Models [25.606641582511106]
  We propose a novel dataset that integrates multi-image inputs with safety Chain-of-Thought (CoT) labels as fine-grained reasoning logic to improve model performance.<n>Our experiments demonstrate that fine-tuning InternVL2.5-8B with MIS significantly outperforms both powerful open-source models and API-based models in challenging multi-image tasks.
  arXiv  Detail & Related papers  (2025-01-30T17:59:45Z)
• MLLM-as-a-Judge for Image Safety without Human Labeling [81.24707039432292]
  In the age of AI-generated content (AIGC), many image generation models are capable of producing harmful content.<n>It is crucial to identify such unsafe images based on established safety rules.<n>Existing approaches typically fine-tune MLLMs with human-labeled datasets.
  arXiv  Detail & Related papers  (2024-12-31T00:06:04Z)
• Safety Without Semantic Disruptions: Editing-free Safe Image Generation via Context-preserving Dual Latent Reconstruction [88.18235230849554]
  Training multimodal generative models on large, uncurated datasets can result in users being exposed to harmful, unsafe and controversial or culturally-inappropriate outputs.<n>We leverage safe embeddings and a modified diffusion process with weighted tunable summation in the latent space to generate safer images.<n>We identify trade-offs between safety and censorship, which presents a necessary perspective in the development of ethical AI models.
  arXiv  Detail & Related papers  (2024-11-21T09:47:13Z)
• Cross-Modality Perturbation Synergy Attack for Person Re-identification [66.48494594909123]
  Cross-modality person re-identification (ReID) systems are based on RGB images.<n>Main challenge in cross-modality ReID lies in effectively dealing with visual differences between different modalities.<n>Existing attack methods have primarily focused on the characteristics of the visible image modality.<n>This study proposes a universal perturbation attack specifically designed for cross-modality ReID.
  arXiv  Detail & Related papers  (2024-01-18T15:56:23Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.