The Data Underminer Project (DUMP)

A (very) temporary place for updates on the project for all interested parties. We will soon have a readthedocs available, which will supercede this site.

The overall project goal is to improve consumer privacy by altering the market for personal data online. We have created a technology that reduces the quality of data collected adversarially online beyond what is gained in increased data volume, so that good privacy practices may become a source of competitive advantage.

For this purpose, part of the project aims to increase transparency for adversarial collection, in two ways: first, by measuring when user email addresses are shared with third parties; and second, by measuring when a website attempts to track visitors against their will (we use a traitor tracing scheme and a browser fingerprinting detection scheme, respectively). A reputation score is then computed that combines these measurements.

Reputation is used to increase accountability. Companies with poor reputations---those who share your email widely, and track you via your browser fingerprint---are more likely to be targeted by adversarial privacy enhancing technologies (APETs) running on our users' machines. An example APET is an automated version of AdNauseum that visits a site and clicks at random on the advertisements displayed, reducing the utility of click data for inferring user preference; however, the system is designed to support community contribution of plug-in adversarial PETs beyond just this one.