Semantic-based Policy Composition for Privacy-demanding Data Linkage

Record linkage can be used to support current and future health research across populations however such approaches give rise to many challenges related to patient privacy and confidentiality including inference attacks. To address this, we present a semantic-based policy framework where linkage privacy detects attribute associations that can lead to inference disclosure issues. To illustrate the effectiveness of the approach, we present a case study exploring health data combining spatial, ethnicity and language information from several major on-going projects occurring across Australia. Compared with classic access control models, the results show that our proposal outperforms other approaches with regards to effectiveness, reliability and subsequent data utility.