Recent work has demonstrated that by monitoring the Real Time Bidding (RTB) protocol, one can estimate the monetary worth of different users for the programmatic advertising ecosystem, even when the so-called winning bids are encrypted. In this paper we describe how to implement the above techniques in a practical and privacy preserving manner. Specifically, we study the privacy consequences of reporting back to a centralized server, features that are necessary for estimating the value of encrypted winning bids. We show that by appropriately modulating the granularity of the necessary information and by scrambling the communication channel to the server, one can increase the privacy performance of the system in terms of K-anonymity. Weve implemented the above ideas on a browser extension and disseminated it to some 200 users. Analyzing the results from 6 months of deployment, we show that the average value of users for the programmatic advertising ecosystem has grown more than 75% in the last 3 years.
Download