Understand and Measure Privacy

In individual pursuits of personalized service and other functionalities, people disclose personal and private information by trusting certain online sites and services. Scholars often assume that such trust is based on a careful assessment of the benefits and risks of disclosing information online. This project departs from such an assumption and investigates the possibility that decision-making about online information disclosure is not systematic, but rather based on cognitive heuristics (or mental shortcuts) triggered by cues in the interaction context.

The project develops new technologies for continual, web-scale measurement and rapid defenses against emerging threats to web privacy and security arising from third-party tracking. It draws from the fields of web security, systems, measurement, statistics, and machine learning. The outputs of this project will enable website administrators to find and fix a large class of privacy and security problems. They will help improve existing browser privacy tools.

This project closely examines data aggregation to understand what types of aggregation are normatively and descriptively important to individuals and how do different types and degree of aggregation impact individual trust. This proposed research would advance knowledge and understanding within the study of big data, trust, and business ethics. Initial investigations into data aggregation have been technical to ensure accuracy and diminish unwanted bias.

Rapid advances in location based applications are leading to increased concern about location privacy. Current mobile operating systems only provide users with rudimentary location access controls - either to block or allow location access - which are inadequate and inefficient in mitigating privacy threats. Most existing location obfuscation mechanisms are based on syntactic privacy models that do not consider mobility and are hence vulnerable to inference attacks.

This project takes a new approach to problems involving sensitive data, by focusing on rigorous mathematical modeling and characterization of the value of private information. By focusing on quantifying the loss incurred by affected individuals when their information is used -- and quantifying the attendant benefits of such use -- the approaches advanced by this work enable concrete reasoning about the relative risks and rewards of a wide variety of potential computations on sensitive data.