Anthony Chavez (GoogleVP, PM, Android Security and Privacy Division) posted a blog on February 16, Introducing the Privacy Sandbox on Android[1].

According to the article:

  • 90% of apps on Google Play are free, largely because of digital advertising
  • A good app ecosystem needs to be good for users, developers and business models
  • Currently, we are planning to use the Privacy Box solution already practiced on the Web to synchronize the solution on the Android platform
  • But some details need to be worked out with the industry, with the first beta release planned for late 2022

Android Privacy SandBox technology solution

Google has planned and implemented the Privacy Sandbox technology solution for Chrome for many years, and now hopes to port it to Android as well.

In theory, the Privacy Sandbox solution could be used with all three party SDKS, but for now the focus will be on governance of the advertising SDK, which is the focus area for using users’ private data.

Through reading official technical documents [2], it can be found that currently on Android platform, solutions mainly include the following four directions:

  • SDK Runtime
  • Topics
  • FLEDGE on Android
  • Attribution Reporting

SDK Runtime

An App released to the market usually contains a large number of three-party SDKS, such as the login SDK, the registration SDK, the PUSH SDK, etc., and most of them also include the advertising SDK.

These SDKS share user data and permissions with apKS (hosts), which may pose a threat to user privacy.

The SDK Runtime solution separates App running from SDK running time, and manages the RUNNING environment of SDK separately.

In this case, SDK no longer enjoys the same permissions and data acquisition ability as APK, and the behavior of obtaining system private data is controlled by the host APK.

The app market also needs to provide support.

The app marketplace should not only distribute apps, but also support SDK upload, approval, and distribution. The App declares the name and version of the SDK that it depends on in the manifest file. When downloading and installing the App, it also downloads the SDK “installation package” for installation.

Topics

Google provides a set of apis for retrieving user preferences

  • Google uses machine learning to identify users’ preferences based on their list of installed apps
  • Google recommends not using the topic cache and fetching it through the interface every time
  • Different apps may get different topics
  • Every time a user installs a new application, a new topic may be generated
  • When a user uninstalls an application, the related topic does not die immediately and lasts for a period of time

FLEDGE on Android

FLEDGE: First Locally-Executed Decision over Groups Experiment

FLEDGE addresses the problem scenario of APP to APP advertising behavior, rather than users’ personal interests. It’s hard to look at the literal meaning, but look at the example.

A is an e-commerce application. If users have unpurchased goods in A’s shopping cart, they need to display advertisements in B. At this point, Google provides two sets of apis to achieve this function, avoiding the sharing or uploading of sensitive data to the server:

  • Custom Audience API
  • Ad Selection API

The logic behind advertising also involves bidding. In the Ad Selection API, the sell-side is directed to place the Selection logic on the side.

Attribution Reporting

Attribution report solves the problem of attribution transformation analysis based on cross-end user identifiers.

The Attribution Reporting API includes:

  • Attribution of report
  • Event-level optimization recommendation reporting based on machine learning
  • Reports of invalid traffic and advertising fraud

The domestic environment

In China, in recent years, government departments (including the Ministry of Industry and Information Technology of the Cyberspace Administration of China) have imposed strict and frequent regulations on Internet application compliance. The head and waist apps on the market were reviewed frequently, and those that failed to comply and complete rectification within the specified time were forced to be removed from the shelves.

It is understood that this year (2022) will also start to focus on SDK compliance issues.

Most of the proposed regulations focus on the acquisition, use and processing of data. If Google’s Android Privacy Sandbox can be successfully introduced to domestic manufacturers, major app stores and developers of head companies cooperate, and laws and regulations are supplemented by constraints, it is believed that the governance of Privacy will be greatly improved.

Ref

  1. Introducing the Privacy Sandbox on Android↩︎
  2. Privacy Sandbox on Android ↩ ︎