Three former Berkeley IEOR/ORMS undergraduate students, Liangyuan Na (B.A. Operations Research & Management Science ’18), Cong Yang (B.S. Industrial Engineering & Operations Research ’18), and Chi-Cheng Lo (B.S. Industrial Engineering & Operations Research ’18) are finalists in the INFORMS Undergraduate Operations Research Prize Competition. The students were advised by IEOR assistant professor Anil Aswani, and their work “Feasibility of Reidentifying Individuals in Large National Physical Activity Data Sets From Which Protected Health Information Has Been Removed With Use of Machine Learning” was published in December 2018 in JAMA Network Open and has been viewed more than 15,000 times since its publication.
The research investigated whether it was feasible to use machine learning algorithms to identify individuals whose protected information had been removed from aggregate datasets of physical activity data.
Currently, companies that track users’ physical activity data through applications and devices can remove individually identifying characteristics from datasets of physical activity data and can then sell the data, make it publicly available or use it for their own commercial purposes. Because they are removing individually identifying information such as name, age, gender, etc. from these datasets, companies can claim that they are respecting individual privacy.
However, by using publicly available data provided by the National Health and Nutrition Examination Survey (NHANES) Na, Yang, and Lo were able to successfully utilize random forest and linear support vector machine algorithms to pair physical activity data with demographic data to predict which record belonged to each individual in the survey. This means that it may be feasible for institutions to purchase physical activity data with identifying information removed and pair with other demographic data to reidentify individuals in physical activity datasets.
The researchers will present their research at the INFORMS Annual Meeting in Seattle in October, 2019.