33 Bits of Entropy

I’m visiting the Bay Area for a couple of weeks. (I’m likely to graduate next Summer and looking to move here.) I gave a talk about my work at Stanford today. It went great, I got lots of feedback and met many cool people. Thanks everyone who showed up!

I will be visiting the Research and Search groups at Microsoft on Monday. I’m also speaking at PARC on Tuesday (11-12) and (tentatively) Berkeley on Wednesday. A couple of other people invited me to give talks at their respective organizations today; I will update once I schedule those.

Here’s the talk abstract:

Privacy and Anonymity in a World of Interconnected Data

The new Web economy relies on the collection of personal data on an ever-increasing scale. Data is collected about our tastes, purchases, searches, browsing history, friendships and relationships, health
history, genetics and so forth. The aggregated datasets are not stationary: they shared with advertisers, marketers and researchers for business reasons. Nor does each such dataset exist in isolation:
it contains implicit or explicit references to other datasets. Unsurprisingly, this has led to a host of privacy issues.

In this talk, I survey the different types of data that are being collected and shared, and propose theoretical models for analyzing privacy in such datasets. Next, I discuss the subtle relationship between anonymity and privacy and present a few related techniques for de-anonymizing large datasets, accompanied by the results of experiments. Finally, I will touch upon the broader threats to privacy arising from these techniques and discuss possible solutions, which, out of necessity, will have a non-technological dimension.