Google officially explains how Android’s Private Compute Core works

[ad_1]

Google introduced several new privacy and security settings with Android 12, including new switches to enable/disable the camera and microphone, indicators to show when the camera and microphone were in use, a new privacy dashboard, and more. While the company has provided detailed explanations for these features, it has only offered a brief overview of Android’s new Private Compute Core.

At last year’s I/O developer conference, Google revealed that the Private Compute Core was a secure partition for processing sensitive user data on the device, similar to those used for passwords and biometrics. The company added that it would be used for several AI-powered features, such as Live Caption, Now Playing and Smart Reply, but didn’t detail how that worked. This led us to speculate that it uses an Android VM dubbed a “microdroid” until a Googler shared some information late last year and revealed that the Private Compute Core was unrelated to running virtual machines. Google has finally offered an official explanation, revealing exactly what Android’s Private Compute Core does and how it works.

In a new blog post, Google explains that the Private Compute Core (PCC) “is a secure, isolated data processing environment within the Android operating system that gives you control over the data within, such as deciding whether , how and when it is shared with others. This way, PCC can enable features like Live Translate without sharing continuous tracking data with service providers, including Google. PCC is part of Protected Computing, a toolkit of technologies that transform how, when and where the data is processed to technically guarantee its privacy and security.”

How Android’s Private Compute Core works

Android’s Private Compute Core essentially keeps sensitive data private for features like Live Translate, Now Playing, and Smart Reply from other subsystems. To do this, Google uses techniques such as Interprocess Communications (IPC) links and isolated processes. These techniques are included in the Android Open Source Project and can be controlled by publicly available surfaces such as the Android Framework APIs.

Overview of the Android Private Compute Core architecture

As Google continually updates its AI capabilities, it also needs to keep the machine learning models that run inside the CCP up to date. To do this without compromising sensitive data, Google leverages federated learning and analytics and monitors network calls to improve model performance using Private Compute Services. Private Compute Services provide a privacy-preserving bridge between Private Compute Core and the cloud, allowing Google to deliver up-to-date AI models and other updates to sandboxed machine learning capabilities over a secure path. Communication between sandboxed features and Private Compute Services is via open source APIs that strip personally identifiable information from the data.

For added transparency, Google has released a technical white paper detailing the data protections in place to keep PCC data confidential, its processes and mechanisms, and privacy framework diagrams for continuous sensing features. Additionally, the company recently open sourced Private Computing Services, allowing the Android community to independently inspect the code that controls data handling and egress policies.

[ad_2]