Paper accepted at BDAA' 2025 and SNAMS 2025
12 September 2025
The paper A Guide to Feature-Preserving Pseudonymization of Profile Pictures authored by Yeong Su Lee, Hendrik Bothe, and Michaela Geierhos was accepted at the First International Conference on Big Data & Applications (BDAA' 2025). This paper covers the pseudonymization of profile images. Unlike existing approaches that focus on text and structured data, profile images can directly identify users. Their pipeline uses FaceNet and DeepFace to extract facial attributes, such as age, gender, and expression. It then formats these attributes as JSON and converts them to descriptive text using Llama 3.3. A multimodal model called Janus then uses this text to generate synthetic, identity-free profile images that retain facial features. All processing is done locally to ensure compliance with the GDPR and avoid data exposure. These pseudonymized images support research and machine learning tasks while protecting privacy.
More about this paper: https://bdaa-conference.com/
Moreover, the contribution A Mixed-Methods Approach to Pseudonymizing Users’ Social Media Profile Data of the same authors was accepted at the 12 International Conference on Social Networks Analysis, Management and Security (SNAMS2025). The primary goal of this research is to address the challenge of preserving privacy while maintaining data utility for downstream applications, such as machine learning. To strike this balance, we propose replacing personal data with context-aware pseudonyms to ensure minimal loss of data quality. Our methodology involves applying lightweight encryption techniques to non-semantic entities, such as user IDs or usernames, while identifying and replacing semantic entities, such as names of people and geographic information, with context-coherent pseudonyms.
More about this paper: https://emergingtechnet.org/SNAMS2025/index.php
Image source: AdobeStock/stardadw007 (Generated with AI)
