Polisis - Automated Analysis and Presentation of Privacy Policies Using Deep Learning

Explore an innovative framework for automated privacy policy analysis in this 25-minute conference talk from USENIX Security '18. Discover Polisis, a scalable tool that utilizes deep learning to comprehend and present complex privacy policies. Learn how a privacy-centric language model, built with 130,000 policies, and a novel hierarchy of neural-network classifiers enable multi-dimensional queries on natural language privacy policies. Examine two practical applications: an automated privacy icon assignment system achieving 88.4% accuracy, and PriBot, the first freeform question-answering system for privacy policies. Gain insights into how these tools can benefit companies, users, researchers, and regulators in navigating the complexities of privacy policies, and understand the potential impact on improving transparency and user comprehension in data collection and sharing practices.