Big Privacy Symposium: Does Publishing a Predictive Model for Precision Medicine Put Patient Privacy at Risk?
Symposium Video
In the context of Big Data, privacy is often thought of as an issue that is managed by de-identifying datasets. However, what are the privacy concerns associated with publishing predictive computational models that are the product of the analysis of confidential datasets. In this talk, Dr. Matt Fredrikson discusses how computational models can potentially be exploited to uncover information from confidential datasets. To prevent this from happening there are precautions that can be taken before publishing predictive models. However, research from Dr. Fredrikson's group shows that the current methods sacrifice either the utility of the model or the confidentially of the data used to create the model.