After a year, Google details how Private Compute Core works

With the launch of Android 12 Last year, Google introduced a new privacy tool that first appeared in Android 12 Beta 2 called Private Compute Core (CPC).

For some time this component However, Android has remained wrapped up in the mysterywhich has given rise to a number of speculations since Google has never revealed how it works in detail, but now it was published a post on his blog that explains exactly how it works.

Follow AndroidWorld on News

To the Google I/O of 2021, the Mountain View house has in fact unveiled the Private Compute Core, describing it as one secure partition for the processing of User data sensitive about device, similar to those used for password hey biometric data. Google then revealed that it would be used for several AI-powered features, such as Live Caption, Now Playing And Smart Reply, but without specifying how it worked.

This led us to imagine that it was a virtual machinebut then an employee of the company had released some information, revealing that it is actually a new type of sandboxes, which allows you to run apps, processes And low-level system modules in isolation.

Now, a year later, Google itself offers one official explanation.

What Private Compute Core does

According to the company’s words, the CCP it’s a processing environment some data safe and isolated within the Android operating system that puts you in control of the data [al suo] internal.

This allows you to decide if, how and when something is shared with the otherswhich allows the CCP to enable features like Live Translate without share data ongoing tracking with service providers, including Google.

Finally, CCP is part of Protected Computinga set of technologies that transforms how, when and where data is processed to technically guarantee its privacy and security.

Therefore Private Compute Core it is an isolated “space” on our phone that allows us to use some Android functions such as Live Translate, Now Playing And Smart Reply without for this we have to share sensitive data with Google, or others.

But how does it work?

How Private Compute Core works

The explanation is quite technicaland is also the subject of a industry articlebut fascinating nonetheless.

Let’s imagine a closed box, the CCP exactlywithin which we can execute applications And processes. Among them there are also models of machine learning used by the apps described above, which must be constantly updated in order to function properly.

So they have to communicate with i server by Google. But to prevent them from sharing even sensitive data, the Private Compute Services (PCS), an app, intervenes stand alone which works from bridge between the CCP and the clouds, enabling Google to provide updated AI models and other feature updates to machine learning sandboxed on a safe path.

This communication is possible thanks to APIs open source that basically they remove identifying information from the data. Below you can find a explanatory diagramthe important part of which concerns precisely the central block, the PCC, and how this communicates with the outside world through the PCS and the APIs of the Android framework.

Overall, Google uses techniques included in theAndroid Open Source Project (AOSP), such as IPC links (Interprocess Communications) e isolated processeswhich can be precisely controlled by the API.

To ensure the maximum transparencyin addition to the above article the Mountain View home has also made open source i Private Computing Services, which allows the community to audit the code and check for anomalies.