Meta Promises to Upgrade Ray Ban Smart Glasses with "Private Processing"

Meta announced in a blog post that its notorious smart glasses will soon use its Private Processing to protect your data as it's being processed in the cloud.

Meta's Ray-Ban glasses have brought widespread criticism at basically all levels of the device.

They were caught sending sensitive recordings to workers for AI training purposes, including "bank details, sex and naked people." The workers needed to label the data so Meta can train their AI models on it.

It was apparently impossible to opt out of at the time as well.

The glasses went on to become infamous for creeps filming people in public without their permission.

The glasses feature a light that's meant to indicate when the camera is active, called the "capture LED." The light is always on and blinks when it's capturing.

People looking to film others surreptitiously tried to disable the camera, and even made businesses out of it.

Meta released an update to disable the camera if tampering was detected. It can also detect if the LED was blocked without being destroyed or tampered with.

They also removed ads and marketplace listings for people offering to tamper with the light.

Despite their efforts, it's still a camera and microphone system attached to your face that can record at any time. Creepy people looking to invade the privacy of others continue to make use of the camera.

This latest feature promises to provide similar protections as on-device AI in the cloud. Their Private Processing was originally implemented in WhatsApp to provide protection to the built-in Meta AI features in the app.

The technology is Meta's implementation of confidential computing. Instead of being processed in a way that's transparent to the operating system and CPU, your query is sent to an isolated environment called a Trusted Execution Environment (TEE).

The TEE is not accessible to the operating system or the rest of the CPU. The memory it uses is encrypted as well "under a key held by dedicated security hardware on the chip."

The feature is similar to features offered by other companies with their AI products such as Apple's Private Cloud Compute and Google's Private AI Compute, meant to "extend your device" to the cloud.

It's important to note that even though confidential computing does offer some protection, ultimately the data must still be decrypted at some point so that the processor can perform operations on it.

There have been vulnerabilities in confidential computing in the past that allowed attackers to access confidential data.