PET Award for making privacy policies easier to read
The research generated a chatbot to help users sift through important details in privacy policies.
A paper co-authored by Kevin and Nancy O’Connor Professor of Computer Science Kang G. Shin has been recognized for its outstanding contribution to privacy technology. The paper, presented at the 2018 USENIX Security Conference and titled “Polisis: Automated analysis and presentation of privacy policies using deep learning,” earned the Caspar Bowden Award for Outstanding Research in Privacy Enhancing Technologies to be presented at PETS 2019 in Sweden. The paper was co-authored with Prof. Florian Schaub from the School of Information and researchers from the Swiss Federal Institute of Technology Lausanne and University of Wisconsin-Madison.
Privacy policies are the primary channel through which companies inform users about their data collection and sharing practices, but as most users know these policies can be long and difficult to comprehend. Motivated by the security implications of these documents, the team developed an automated framework for analyzing privacy policies called Polisis. The tool enables scalable, dynamic, and multi-dimensional queries on natural language privacy policies.
Polisis was built with 130,000 privacy policies, and a hierarchy of neural-network classifiers that can account for both high-level and fine-detail features of privacy practices. The framework was used to create two applications: a querying application that can assign privacy icons to a given policy, and PriBot, the first freeform question-answering chatbot for privacy policies. Their tests demonstrated that PriBot can produce a correct answer among its top-3 results for 82% of the test questions.