Endpoint User Profile Control

This research is supported by the National Science Foundation.

Motivation and Approach

The Internet has transformed from a small non-profit network into a gigantic infrastructure that creates revenues measured in billions of dollars. Unfortunately, this has created an unprecedented pressure on the end user's privacy, because a user's profile is used to determine which personalized content (e.g., ads, search results, recommendations, etc.) will be served to the given user. It is not a secret that almost every browsing click we make is collected by one or more of numerous information trackers and aggregators associated with various online services. Many argue that stakes are much higher and go far beyond simple content personalization. In particular, they argue that the current practices create the so-called "filter bubble" effect, which has an even more profound impact on the society as a whole, even on the future of democracy. Necessarily, users are deeply concerned by the increasing levels in which their personal information is collected, stored, and used in various online personalization contexts. According to a recent poll, 72 percent of Americans are worried that their online behavior is being tracked and profiled by companies.

We propose endpoint user profile control as a comprehensive approach to the above personalization-induced problems. In our approach, the user has the ability and means, which we will develop, to explicitly define and implicitly control its profile at all possible trackers at once by leaving synthetic controlled online footprints. Our system allows users to submit their preferences to all possible trackers they encounter, granting the user a single point of control over their profiles. Since the system functions by generating direct traffic, it stands to work with all trackers. As traffic is generated directly from the machine which the user uses to access online services, our approach is able to function with any method of tracking, even those which use techniques beyond simple cookie interactions. Furthermore, our system takes advantage of existing infrastructure and systems, requiring action only from users.

This project aims to develop endpoint user profile control methodologies and tools in multiple domains. In particular, we focus on (i) enabling endpoint user profile control for the Web advertising domain, (ii) endpoint user profile control for search engines and information aggregators, and (iii) quantifying and containing explicit user privacy leaks from an endpoint.



  • Oak: User-Targeted Web Performance
    M. Flores, A. Wenzel, and A. Kuzmanovic
    In IEEE ICDCS '17, Atlanta, Georgia, June 2017.
    [ To appear ]

  • Messup: Protecting User Identities in Mobile Traffic
    N. Xia and A. Kuzmanovic
    Submitted for publication.
    [ To appear ]

  • Synthoid: Endpoint User Profile Control
    M. Flores and A. Kuzmanovic
    In IEEE/WIC/ACM International Conference on Web Intelligence, Warsaw, Poland, August 2014.
    [ .pdf | .pdf.gz ]

  • Mosaic: Quantifying Privacy Leakage in Mobile Networks
    N. Xia, H. Song, Y. Liao, M. Iliofotou, A. Nucci, Z. Zhang, and A. Kuzmanovic
    In Proceedings of ACM SIGCOMM 2013, Hong Kong, China, August 2013.
    [ .pdf | .pdf.gz ]

  • Searching for Spam: Detecting Fraudulent Accounts Via Websearch
    M. Flores and A. Kuzmanovic
    Proceedings of the 14th International Conference on Passive and Active Measurement, Hong Kong, China, March 2013.
    [ .pdf | .pdf.gz ]