Android 12 adds AI and machine learning with Private Compute Core but keeps your data secure

Google has gone to great lengths to secure the Android platform, while at the same time creating a platform that is intelligent, intuitive and predictive. To continue to push the boundaries of smartphone technology, Google relies heavily on machine learning. But instead of offloading ML and artificial intelligence to the cloud, the developers have opted to keep it on-device. With this, Android is better able to anticipate a user’s next move or learn their habits (by recognizing and repeating patterns).

It is that functionality that led the developers to bring the computation of ML and AI onto the devices. Imagine the security issues that would be in play if that service was offloaded to a third-party compute platform. In today’s world of constant threat, that would not fly. For that, Google developed the Private Compute Core, which arrived with the release of Android 12. 

SEE: Electronic communication policy (TechRepublic Premium)

As of now (Android 12, beta 4), the Private Compute Core works with three features:

• Live Caption
• Smart Reply
• Now Playing

Prior to Android 12, the AI powering the above features was already found on the platform. It wasn’t until Android 12, however, that the feature was given a dedicated location within Settings. If you go to Settings | Privacy | Private Compute Core, you’ll find all of the available options (Figure A).

Figure A

Outside of the features Private Compute Core will bring to your device, one of the more important things it does is store data separate from all apps and services, while making that data available to the operating system when required. For example: When you watch a video with Live Caption, the service that listens to the audio and translates it to real-time, readable captions cannot expose the audio to anyone or any other app. With Private Compute Core at work, that data is processed within its own sandbox, away from everything else, and then the captions are presented for your viewing.

Private Compute core will make it possible for Google to develop and release new features while keeping data safe, private and local.

How Android’s Private Compute Core works

The idea behind Private Compute Core is simple: Everything will happen within a privileged space, similar to how separate partitions are used for passwords and biometric data. Instead of housing credentials, the Private Compute Core partition will hold data used for ML (such as data for Smart Reply or the Now Playing feature to identify songs).

SEE: How to enable themed icons in Android 12 (TechRepublic) 

You might think that, because of the name Private Compute Core uses a dedicated chip. It does not. Instead, the Private Compute Core dedicated partition resides entirely in software. The reason for this design is so that Private Compute Core can be deployed across a wide range of Android hardware. Had Google gone the dedicated chip route, Private Compute Core would be limited to Pixel devices only.

How to clear learned data in Private Compute Core

Google was smart to add the ability to clear all learned data. However, this does place a modicum of responsibility on the shoulders of the end-users. Should you expect some bit of data (used by the ML/AI system) has been compromised or used for purposes outside the stated norm for the platform, you can head to Settings | Privacy | Private Compute Core and tap Clear Data. A popup will appear, where you can select the time frame from which you’d like data to be cleared (Figure B).

Figure B

I suspect more and more features will be coming to Private Compute Core. It might not happen until the next iteration of the platform, but you can be certain that Google will continue leaning heavily on ML and AI to power Android even further into the future. Having siloed the data used by ML/AI, Google can better guarantee the safety of that data going forward.

5G and Mobile Enterprise Newsletter

5G networks and devices, mobile security, remote support, and the latest about phones, tablets, and apps are some of the topics we’ll cover.
Delivered Tuesdays and Fridays

Sign up today

Also see