Evaluation of Hardening Techniques for Privacy-Preserving Record Linkage
Proceedings of the 24th International Conference on Extending Database Technology (EDBT), pp. 289-300
Futher information: https://doi.org/10.5441/002/edbt.2021.26
Privacy-preserving record linkage aims at integrating person-related data from different sources while protecting the privacy of individuals by securely encoding and matching quasi-identifying attributes, like names. For this purpose Bloom-filter-based encodings have been frequently used in both research and practical applications. Simultaneously, however, weaknesses and attack scenarios were identified emphasizing that Bloom filters are in principal susceptible to cryptanalysis. To counteract such attacks, various encoding variants and tweaks, also known as hardening techniques, have been proposed. Usually, these techniques bear a trade-off between privacy (security) and the linkage quality outcome. Currently, a comprehensive evaluation of the suggested hardening methods is not available. In this work, we will therefore review and categorize available Bloom-filter-based encoding schemes and hardening techniques. We also comprehensively evaluate the approaches in terms of privacy (security) and linkage quality to assess their practicability and their effectiveness in counteracting attacks.