Facebook announced that they are investing in research supporting the furthering of Privacy Enhancing Technology, through academics, global organizations and developers.
It’s no surprise that Facebook is taking a heavy interest in Privacy-Enhancing Technologies after the changes stemming from iOS14 have created massive losses in data for advertisers. The goal in investing in these technologies is to identify better methods of tracking which will maintain privacy while feeding anonymized and aggregated data back to Facebook, bettering advertiser outcomes.
Privacy-Enhancing Technologies and Facebook Ads
Privacy-Enhancing Technologies (PET) are technologies that can minimize the amount of data processed with the goal of protecting personal information.
Facebook shared examples of Covid contact tracing and sending electronic payments to illustrate the ways that PET can be used to track need-to-know information while still protecting personal data.
There are three primary PETs that Facebook is investing in, which include: Multi-Party Computation, On-Device Learning, and Differential Privacy. Let’s delve into each of those.
Multi-Party Computation (MPC) & Private Lift Measurement
Facebook has been testing a solution called Private Lift Measurement, which uses multi-party computation (MPC) to help advertisers understand performance while keeping consumer data private.
MPCs are used to calculate outcomes using data sources from multiple parties. For instance, in Facebook’s use-case, this type of reporting is used to combine ad engagement data from one party and purchase data from another.
Facebook expects this measurement to be available to all advertisers next year but for now, has open-sourced the framework so that any developer can create privacy-centric measurement products using MPC.
On-Device Learning is just as it sounds – tracking that lives in the individual device, which then trains an algorithm about particular habits and likely future behaviors.
For instance, Facebook gives the example that if people click on exercise equipment also tend to buy protein shakes, then on-device learning would detect those patterns without sending that individual data to the cloud.
This sounds somewhat similar to what Google Chrome is trying to accomplish with FLoC by keeping browsing data within the individual browser.
Last but not least, differential privacy calculates the noise in a data set. It anonymizes the data by making small changes to it, to make it more difficult to know exactly who took a particular action.
This technology is often used for public research for that reason. Differential privacy can be used on its own or with other privacy-enhancing technologies.
When Can We Expect To See Changes?
Facebook didn’t give an exact timeline for when changes were expected but they did mention that the initiative is a multi-year effort. Presumably, they’ll begin testing things within that time frame but advertisers may not see major changes in the immediate future.