mlsanityData to 14 Oct 2026

Protecting Your Video Content: Disrupting Automated Video-based LLM Annotations

arXiv 2503.21824 ↗First posted 26 Mar 2025cs.CV cs.CRcited by 5

On the map: LLM safety and security → Watermarking generated content and models

Abstract (arXiv)

Recently, video-based large language models (video-based LLMs) have achieved impressive performance across various video comprehension tasks. However, this rapid advancement raises significant privacy and security concerns, particularly regarding the unauthorized use of personal video data in automated annotation by video-based LLMs. These unauthorized annotated video-text pairs can then be used to improve the performance of downstream tasks, such as text-to-video generation. To safeguard personal videos from unauthorized use, we propose two series of protective video watermarks with imperceptible adversarial perturbations, named Ramblings and Mutes. Concretely, Ramblings aim to mislead video-based LLMs into generating inaccurate captions for the videos, thereby degrading the quality of video annotations through inconsistencies between video content and captions. Mutes, on the other hand, are designed to prompt video-based LLMs to produce exceptionally brief captions, lacking descriptive detail. Extensive experiments demonstrate that our video watermarking methods effectively protect video data by significantly reducing video annotation performance across various video-based LLMs, showcasing both stealthiness and robustness in protecting personal video content. Our code is available at https://github.com/ttthhl/Protecting_Your_Video_Content.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Placed in Watermarking generated content and models, one of 12 directions in LLM safety and security. Builds on 41 of its 52 references that are in the map. First posted on arXiv 26 Mar 2025.

Builds on

41 of 52 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Video Watermarking: Safeguarding Your Video from (Unauthorized) Annotations by Video-based LLMs2 Jul

Other references · 40

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2020Language Models are Few-Shot Learnerscited by 32,419
  3. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  4. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  5. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  6. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  7. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  8. 2023Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understandingcited by 1,471
  9. 2023AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuningcited by 1,433
  10. 2021Frozen in Time: A Joint Video and Image Encoder for End-to-End Retrievalcited by 1,177
  11. 2023Video-ChatGPT: Towards Detailed Video Understanding via Large Vision and Language Modelscited by 1,062
  12. 2024Latte: Latent Diffusion Transformer for Video Generationcited by 546
28 more
  1. 2023VideoCrafter1: Open Diffusion Models for High-Quality Video Generationcited by 535
  2. 2019On Evaluating Adversarial Robustnesscited by 447
  3. 2024OpenVid-1M: A Large-Scale High-Quality Dataset for Text-to-video Generationcited by 341
  4. 2023EvalCrafter: Benchmarking and Evaluating Large Video Generation Modelscited by 340
  5. 2023Visual Adversarial Examples Jailbreak Aligned Large Language Modelscited by 328
  6. 2024PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioningcited by 322
  7. 2023On Evaluating Adversarial Robustness of Large Vision-Language Modelscited by 284
  8. 2023Vid2Seq: Large-Scale Pretraining of a Visual Language Model for Dense Video Captioningcited by 263
  9. 2023VTimeLLM: Empower LLM to Grasp Video Momentscited by 255
  10. 2023How Robust is Google's Bard to Adversarial Image Attacks?cited by 183
  11. 2022Untargeted Backdoor Watermark: Towards Harmless and Stealthy Dataset Copyright Protectioncited by 86
  12. 2024Inducing High Energy-Latency of Large Vision-Language Models with Verbose Imagescited by 69
  13. 2022Black-box Dataset Ownership Verification via Backdoor Watermarkingcited by 65
  14. 2024Koala: Key frame-conditioned long video-LLMcited by 57
  15. 2022NICGSlowDown: Evaluating the Efficiency Robustness of Neural Image Caption Generation Modelscited by 50
  16. 2024AnimateDiff-Lightning: Cross-Model Diffusion Distillationcited by 47
  17. 2024One Perturbation is Enough: On Generating Universal Adversarial Perturbations against Vision-Language Pre-training Modelscited by 40
  18. 2020Targeted Attack for Deep Hashing based Retrievalcited by 35
  19. 2024Denial-of-Service Poisoning Attacks against Large Language Modelscited by 35
  20. 2023Fool Your (Vision and) Language Model With Embarrassingly Simple Permutationscited by 25
  21. 2024Adversarial Robustness for Visual Grounding of Multimodal Large Language Modelscited by 23
  22. 2024Deconstructing The Ethics of Large Language Models from Long-standing Issues to New-emerging Dilemmas: A Surveycited by 23
  23. 2022MOVE: Effective and Harmless Ownership Verification via Embedded External Featurescited by 21
  24. 2024PointNCBW: Towards Dataset Ownership Verification for Point Clouds via Negative Clean-label Backdoor Watermarkcited by 20
  25. 2024CLIP-Guided Generative Networks for Transferable Targeted Adversarial Attackscited by 19
  26. 2023Defenses in Adversarial Machine Learning: A Surveycited by 14
  27. 2024Embedding Self-Correction as an Inherent Ability in Large Language Models for Enhanced Mathematical Reasoningcited by 13
  28. 2024Efficient Self-Supervised Video Hashing with Selective State Spacescited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 3; 2026: 2.

5 papers.

  1. 2025HLFormer: Enhancing Partially Relevant Video Retrieval with Hyperbolic Learningcited by 17
  2. 2025VideoGuard: Protecting Video Content from Unauthorized Editingcited by 3
  3. 2025Towards Dataset Copyright Evasion Attack against Personalized Text-to-Image Diffusion Modelscited by 2
  4. 2026VideoSTF: Stress-Testing Output Repetition in Video Large Language Modelscited by 1
  5. 2026Controllable Multi-label Video Safety Detection via Adaptive Tversky Policy Optimizationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Failures to Surface Harmful Contents in Video Large Language Models
  2. 2025VideoEraser: Concept Erasure in Text-to-Video Diffusion Models
  3. 2026Membership Inference Attacks Against Video Large Language Models
  4. 2024Can Watermarking Large Language Models Prevent Copyrighted Text Generation and Hide Training Data?
  5. 2025LexiMark: Robust Watermarking via Lexical Substitutions to Enhance Membership Verification of an LLM's Textual Training Data
  6. 2024Can Watermarked LLMs be Identified by Users via Crafted Prompts?
  7. 2025Can LLM Watermarks Robustly Prevent Unauthorized Knowledge Distillation?
  8. 2026Jailbreaking Multimodal Large Language Models using Multi-Clip Video