[3K5-OS-2b] OS-2

[3K5-OS-2b-03] Privacy Benchmarks in Anonymized healthcare data

〇Hiroaki Kikuchi¹, Koji Chida² (1. Meiji University, 2. Gunma University)

Keywords:personal information, anonymization, privacy

Healthcare data such as lifelog and medical diagnostics are useful for quantifying risks to be suffered in diabetes and other diseases. Insufficient privacy techniques used in anonymizing and synthesizing have risks, e.g., singling out, record likability and inference attacks. To quantify risks in anonymized bigdata, privacy metrics used in data anonymization competition PWS Cups and opensource platform Anonymeter have been proposed so far. In this work, we evaluate risks of use-case in diabetes based on features including age, education level, BMI, physical activities in NHANES datasets (National Health and Nutrition Examination Survey) due to US CDC and discuss the reliable privacy benchmarks.