Elamur Research⌕
☰
home / papers / 2026 / Privacy VITA: a new multilingual and multimodal annotated video corpus to evaluate anonymization systems

Privacy VITA: a new multilingual and multimodal annotated video corpus to evaluate anonymization systems

Paper · 2026 · Jorge Rico, Sofia Contreras, Enrique Manjavacas Arevalo et al.

Summary

Personal data increasingly appears in multimedia, but resources for evaluating anonymization are split by modality: tabular, text, faces or voice. The authors report that no public annotated dataset supports end-to-end anonymization across video, image, text and audio. To fill this gap they built Privacy VITA, a corpus of 516 CC BY YouTube videos (about 1,756 minutes) in eight European languages and several regional variants. Researchers selected the videos by hand, and separate teams screened them for ethical and technical suitability. Annotation ran in two parallel workflows. Native speakers corrected Whisper transcripts, segmented speaker turns and tagged 12 PII entity types in speech. Image annotators drew bounding boxes around identity-revealing objects every 20 frames and added attributes to them. Automatic consistency checks and an independent review team then refined the labels. A pilot test of three PII detectors on the English, Spanish and French subsets showed weak and uneven results across languages and entity types. The corpus gives researchers a shared testbed for multimodal, multilingual anonymization, which matters for GDPR compliance.

Key points

  • Coverage: The 516 videos cover eight languages (Basque, Catalan, English, French, Galician, Italian, Portuguese and Spanish) along with regional variants like Brazilian Portuguese and Canadian French. The videos span 11 domains, with vlogs being the most PII-rich.
  • Annotation scheme: 12 PII entity types in audio (e.g., NAME, GPE, ORG, ROLE, DISEASE, ADDRESS, NUM_BANK) and six visual object classes (e.g., faces, persons, license plates, logos, screens). The corpus contains 6,943 audio entities and more than 100,000 annotated objects.
  • Process: Annotation was done in LabelStudio by 15 native-speaker annotators for audio/text and 10 image annotators. Each item had a single annotator, so no inter-annotator agreement was computed, but senior annotators supervised the work.
  • Baseline: LLM-Guard, Presidio and PrivateAI were tested with entity-level F1 after label normalization. ADDRESS scored 0.00 for every system and language. The best micro-averaged F1 was 0.44 for English (Presidio), 0.36 for Spanish (PrivateAI) and 0.41 for French (Presidio).
  • Distribution: The corpus consists of JSON files keyed to YouTube IDs, with metadata on language, dialect, domain, lighting and resolution. It is available on request and is planned for release under CC-BY-NC-SA.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksComputer Vision and Pattern Recognition · 2,869 worksSocial Sciences · 1,772 worksSociology and Political Science · 577 works

🧭 Topics 2Research subjects and languages the work is about.Multilingual & cross-lingual NLP · 828 worksVision & language · 657 works

Cite

Cite as Jorge Rico, Sofia Contreras, Enrique Manjavacas Arevalo, Maria Viana, Ruben Pérez-Ramón, Maria Luisa Izquierdo, María José Vilella, Jaime Corton, Silvia Rodriguez, Fernando Espinza, Rafael Ginard, César Pérez, Jesús Arias, Richard Cook, Pablo Regodón, Manuel Moyano, Ricardo Heredia, Lara De Santos, Pierre Plaza and Jacqueline González. 2026. Privacy VITA: a new multilingual and multimodal annotated video corpus to evaluate anonymization systems. Proceedings of the Second International Conference on Natural Language Processing and Artificial Intelligence for Cyber Security.

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{rico2026privacy,
  title = {{Privacy VITA: a new multilingual and multimodal annotated video corpus to evaluate anonymization systems}},
  author = {Jorge Rico, Sofia Contreras, Enrique Manjavacas Arevalo, Maria Viana, Ruben Pérez-Ramón, Maria Luisa Izquierdo, María José Vilella, Jaime Corton, Silvia Rodriguez, Fernando Espinza, Rafael Ginard, César Pérez, Jesús Arias, Richard Cook, Pablo Regodón, Manuel Moyano, Ricardo Heredia, Lara {De Santos}, Pierre Plaza and Jacqueline González},
  booktitle = {Proceedings of the Second International Conference on Natural Language Processing and Artificial Intelligence for Cyber Security},
  year = {2026},
  url = {https://aclanthology.org/2026.nlpaics-1.27/}
}

References

26 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Victoria Arranz, Khalid Choukri, Montse Cuadros, Aitor García-Pablos, Lucie Gianola, Cyril Grouin, Manuel Herranz, Patrick Paroubek, and Pierre Zweigenbaum. 2022. MAPA Project: Ready-to-Go Open-Source Datasets and Deep Learning Technology to Remove Identifying Information from Text Documents. In Pro… find
  2. Article 29 Data Protection Working Party. 2014. Opinion 05/2014 on anonymisation techniques. Technical report, European Commission. find
  3. Tadas Baltrušaitis, Chaitanya Ahuja, and Louis-Philippe Morency. 2019. Multimodal machine learning: A survey and taxonomy. IEEE Transactions on Pattern Analysis and Machine Intelligence, 41(2):423–443. find
  4. Barry Becker and Ronny Kohavi. 1996. UCI Machine Learning Repository. https://doi.org/10.24432/C5XW20. — [doi:10.24432/c5xw20](https://doi.org/10.24432/c5xw20) ↗
  5. Adult. DOI: Carole Cadwalladr and Emma Graham-Harrison. 2018. Revealed: 50 million facebook profiles harvested for cambridge analytica. The Guardian. find
  6. Paul Cook, Helena Smith, and Steven Bird. 2019. Multimodal datasets for human communication research: Review and recommendations. Language Resources and Evaluation, 53(2):541–565. find
  7. Khaled El Emam and Fida K. Dankar. 2008. Protecting privacy using k-anonymity. Journal of the American Medical Informatics Association, 15(5):627–637. find
  8. Electronic Privacy Information Center (EPIC). n.d. The privacy act of 1974. https://epic.org/ the-privacy-act-of-1974/. Accessed: 2025-10-09. ↗
  9. European Parliament and Council. 2016. General data protection regulation (gdpr). Official Journal of the European Union, L119:1–88. find
  10. Luciano Floridi and Josh Cowls. 2022. A unified framework of five principles for ai in society. Harvard Data Science Review, 4(1). find
  11. IBM Security. 2021. Cost of a data breach report 2021. Technical report, IBM Corporation. find
  12. David Leslie. 2019. Understanding artificial intelligence ethics and safety. CoRR, abs/1906.05684. — [arXiv:1906.05684](https://arxiv.org/abs/1906.05684) find
  13. Sarina Meyer, Florian Lux, and Pavel Denisov. 2022. Speaker anonymization with phonetic intermediate representations. In Interspeech 2022, pages 4925– 4929. find
  14. Sarina Meyer, Florian Lux, and Ngoc Thang Vu. 2024. Probing the Feasibility of Multilingual Speaker Anonymization. In Interspeech 2024, pages 4448– 4452. find
  15. Michele Panariello, Natalia Tomashenko, Xin Wang, Xiaoxiao Miao, Pierre Champion, Hubert Nourtel, Massimiliano Todisco, Nicholas Evans, Emmanuel Vincent, and Junichi Yamagishi. 2024. The voiceprivacy 2022 challenge: Progress and perspectives in voice anonymisation. IEEE/ACM Transactions on Audio, Sp… find
  16. AJ Piergiovanni and Michael Ryoo. 2020. Avid dataset: Anonymized videos from diverse countries. Advances in Neural Information Processing Systems, 33:16711–16721. find
  17. Ildikó Pilán, Pierre Lison, Lilja Øvrelid, Anthi Papadopoulou, David Sánchez, and Montserrat Batet. 2022. The text anonymization benchmark (tab): A dedicated corpus and evaluation framework for text anonymization. Computational Linguistics, 48(4):1053–1101.
  18. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2023. Robust speech recognition via large-scale weak supervision. In Proceedings of the 40th International Conference on Machine Learning (ICML), Honolulu, Hawaii.
  19. Felix Rosberg, Eren Erdal Aksoy, Cristofer Englund, and Fernando Alonso-Fernandez. 2023. Fiva: facial image and video anonymization and anonymization defense. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 362– 371. find
  20. Andreas Rossler, Davide Cozzolino, Luisa Verdoliva, Christian Riess, Justus Thies, and Matthias Nießner. 2019. Faceforensics++: Learning to detect manipulated facial images. In Proceedings of the IEEE/CVF international conference on computer vision, pages 1–11. find
  21. Sandip Shinde, Arya Pathak, Shreya Mohite, Sanchitsai Nipanikar, and Keyur Pande. 2025. A multimodal anonymization framework for mp4 videos. In 2025 International Conference on Emerging Smart Computing and Informatics (ESCI), pages 1–6. find
  22. Label Studio. 2026. Label studio annotation software. Online. [Visited on 2026-03-13]. find
  23. Verizon Business. 2023. Data breach investigations report 2023. Technical report, Verizon. find
  24. Samuel D. Warren and Louis D. Brandeis. 1890. The right to privacy. Harvard Law Review, 4(5):193–220. find
  25. Zhenyu Wu, Haotao Wang, Zhaowen Wang, Hailin Jin, and Zhangyang Wang. 2022. Privacy-preserving deep action recognition: An adversarial learning framework and a new dataset. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(4):2126–2139. find
  26. Jie Xu, Feng Zhou, Yong Wang, and Qi Liu. 2019. An evaluation framework for data utility and privacy trade-offs. IEEE Access, 7:10372–10383. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/rico-2026-privacy-vita-new-multilingual-multimodal/