Facebook Is Rebuilding Its Ad Tools With A Focus On Privacy

During the apple WWDC developer 2021 festival in June, announcements were made regarding the fact that iOS’ operating system will be more focused towards addressing privacy needs. Google also decided to key into the privacy-focused system by deciding that it will stop selling ads according to a user’s web browser data.

Facebook in the same vein, is not letting the privacy-focus era pass it by. It is presently working on a solution that presents user-privacy as a key subject. This solution is what the social media giant has termed as a Secure multi-party computation-party computation (MPC). This term enables advertisers to understand how ad campaigns are performing, while making sure that neither Facebook, nor the person organizing the ad campaign (the advertiser in this case), can learn any information that does not affect their business.

With the MPC, data is end-to-end encrypted while in transit- this means that neither party can see each other’s data. With this configuration, both parties will gain insight about how an ad is performing without a party having to be in the know of two sets of data. The Multi-Party Configuration will be available on Facebook Business Manager platforms fully next year.

Any privacy-focused solutions presently?

YES. Facebook is currently working on Private Lift Measurement – a solution that uses MPC to help advertisers understand ad performance. What is good about this is that Private Computation Framework is open-sourced, so developers can create their own privacy-centric measurements using MPC.

Away from MPC, Facebook is trying out on-device learning. On-device learning improves a person’s ad experience by processing their data on their device without having to first send the data to a remote server/cloud. With this measure, Facebook will be able to send relevant ads to a Facebook user without having to peep on information about the user’s browsing data or their app choices and preferences.    

To strengthen the efficiency of on-device learning, another measure called differential privacy will be adopted. Differential noise adds “noise” to a data set. The aim is to add small random bits of information to make it harder to know who bought a product for example, after clicking on an ad.

Year on year, Facebook is making effort to build a portfolio of privacy-enhancing technologies (PET). This minimizes the amount of data to be processed so that personal information will not be at risk.

With PETs, Facebook hopes that data will be made anonymous, and more aggregate so that original data will not be compromised. The aggregate data will then be used to run personalized ads.

Speaking to The Verge, Facebook’s Vice President of Product marketing for ads, Graham Mudd said: “we definitely see that (ads) personalization will evolve very meaningfully over the course of the next five years and that investing well ahead of that will benefit all of our customers and enable us to help shape that future state of the ads ecosystem”.