Google FLoC: Design Proposal Project

For CMU course 17-631, my team and I researched Google FLoC and the Privacy Sandbox. We spent the semester learning about its algorithms, where it fails, where it succeeds, public response, and competitor alternatives. A link to our final paper is included below.

We investigate the Federated Learning of Cohorts (FLoC), Google’s current keystone for its plan to replace third-party cookies. Our investigations included in-depth literature reviews as well as conversations with experts on privacy economics, cryptography, and web-tracking. We deliver a discussion on the ongoing debate surrounding FLoC from its announcement to its current trial phases. We ultimately find several faults with FLoC and view it as unsustainable in its current form. Finally, we propose a design for improving FLoC’s privacy guarantees while still providing good advertiser utility.