Chair of Data Science at the SNAMS 2025
2 Dezember 2025
From November 24 to 28, 2025, the 12th International Conference on Social Networks Analysis, Management, and Security (SNAMS2025) took place at the University of Applied Sciences BFI in Vienna. The conference is an interdisciplinary forum where researchers, students, and experts present and discuss new contributions to the analysis, management, and security of social networks, with a particular focus on data-driven methods (data mining, machine learning) and the challenges posed by large, dynamic online platforms such as Facebook, X, and LinkedIn.
The Chair of Data Science was represented by Yeong Su Lee and Hendrik Bothe – speaker and session chair. Their presentation, “A Mixed-Methods Approach to Pseudonymizing Users' Social Media Profile Data,” introduced a method for pseudonymizing data entities in social media datasets. The chosen approach is intended to prevent conclusions about real persons from being drawn, while still containing information that allows meaningful conclusions to be drawn about correlations (e.g., number of persons with an employer in the IT or automotive industry).
The core idea: Profile attributes of the data set are separated into semantic (e.g., names of people, places, organizations) and non-semantic entities (IDs, purely technical information). The semantic entities are replaced in a context-coherent manner. This is done using a locally executed large language model (LLM). Non-semantic entities, on the other hand, are deterministically transformed using a modified Hill cipher. This preserves realistic relationships (family, education, job changes, moves) and, consequently, the evaluability of the data without revealing the identity of individual persons. The method was demonstrated on a Facebook dataset with around 7,700 public profiles.
Image sources: Hendrik Bothe / Yeong Su Lee (RI CODE)
