Preserving Privacy in Multimedia : Text-Aware Sensitive Information Masking for Visual Data
DOI:
https://doi.org/10.32628/CSEIT2410117Keywords:
Text Detection, NLP, ML, Masking, Image Detection, NER, BlurringAbstract
The unauthorised revelation of confidential data has become a source of concern due to the proliferation of multimedia content on the Internet. Unintentionally captured and exposed textual data, including but not limited to personally identifiable information (PII), financial details, and confidential documents, may be present in images and videos when they are being recorded or shared. This study presents an innovative method for concealing text-sensitive information in visual data with the intention of safeguarding privacy without compromising the context and integrity of multimedia content. By utilising cutting-edge text detection algorithms, our approach effectively discerns textual areas present in images and videos. Natural language processing (NLP) methods are subsequently utilised to categories the identified text into sensitive or non-sensitive categories according to predetermined standards. In the case of confidential information, we employ a context-aware concealing strategy that obscures only the pertinent segments while maintaining the visual indicators and encircling context. This approach diverges substantially from conventional pixel-level masking, which frequently obliterates crucial data and impedes interpretability. In pursuit of context-aware masking, we investigate a range of methodologies including semantic-based keyword masking, character-level redaction, and text-region-specific image inpainting. The efficacy of our methodology is assessed across a range of datasets comprising videos and images that contain text types and sensitivity levels that vary. The accuracy of text detection and classification, the impact on user comprehension, and the effectiveness of concealing in preserving privacy and visual quality are all evaluated. The findings of this study carry substantial ramifications for safeguarding privacy across a range of domains, encompassing social networking services, online media sharing platforms, and video surveillance systems. Our approach provides a valuable tool for protecting personal information and upholding privacy rights in the digital age by facilitating the selective masking of sensitive text while preserving the visual fidelity and context of multimedia content.
References
- Amini, A. A., Roussos, A., & Katsaggelos, A. K. (2019). Context-aware privacy preserving for images via semantic region redaction. In Proceedings of the IEEE International Conference on Image Processing (ICIP) (pp. 3671-3675). IEEE.
- Wang, F., Chen, F., Zhang, F., & Liu, X. (2020). Privacy-preserving image redaction using salient object detection and attention-based inpainting. IEEE Transactions on Circuits and Systems for Video Technology, 31(7), 1505-1518.
- Zhou, X., Yao, C., Wen, H., Liu, Y., & Tian, S. (2014). Text detection and recognition in imagery: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 36(7), 1489-1509.
- Hossain, M. S., Hasan, M. A., & Pickering, M. D. (2019). Deep learning-based privacy protection for online social networks. arXiv preprint arXiv:1909.02007.
- El Bouchti, M. A., Foukar, F., & Khelladi, M. T. (2018). De-identification for privacy protection in multimedia content: A survey. Computers & Security, 74, 308-332.
- Jia, X., Gong, N., Liu, X., Xiang, Y., & Zhou, Y. (2019). Privacy-preserving image sharing via homomorphic encryption with secure computation. arXiv preprint arXiv:1905.04465.
- Li, X., Wang, A., Shi, Z., & Chai, T. (2019). Towards privacy-preserving facial expression recognition. In Proceedings of the 2019 ACM Multimedia Conference (pp. 2567-2575).
- K. Menaka and B. Yogameena (2021). Face Detection in Blurred Surveillance Videos for Crime Investigation. In Journal of Physics: Conference Series. DOI 10.1088/1742-6596/1917/1/012024.
- Li, M., Zhang, S., Wang, W., & Feng, W. (2019). A survey of privacy protection techniques for text in images and videos. arXiv preprint arXiv:1903.04517.
- Zhao, J., Xu, Y., Yang, Y., & Wang, Y. (2019, June). Towards context-aware masking of sensitive information in images. In Proceedings of the International Conference on Multimedia (pp. 3098-3102).
Downloads
Published
Issue
Section
License
Copyright (c) IJSRCSEIT

This work is licensed under a Creative Commons Attribution 4.0 International License.