Personal information privacy: what's next?

Official URL: http://ceur-ws.org/Vol-2622/paper5.pdf

In recent events, user-privacy has been a main focus for all technological and data-holding companies, due to the global interest in protecting personal information. Regulations like the General Data Protection Regulation (GDPR) set firm laws and penalties around the handling and misuse of user data. These privacy rules apply regardless of the data structure, whether it being structured or unstructured. In this work, we perform a summary of the available algorithms for providing privacy in structured data, and analyze the popular tools that handle privacy in textual data; namely medical data. We found that although these tools provide adequate results in terms of de-identifying medical records by removing personal identifyers (HIPAA PHI), they fall short in terms of being generalizable to satisfy nonmedical fields. In addition, the metrics used to measure the performance of these privacy algorithms don't take into account the differences in significance that every identifier has. Finally, we propose the concept of a domain-independent adaptable system that learns the significance of terms in a given text, in terms of person identifiability and text utility, and is then able to provide metrics to help find a balance between user privacy and data usability.