WhatsApp has announced the introduction of ‘private processing’, a new technique that enables users to use advanced AI features by taking off the functions of the privacy-conservation cloud server.
It is necessary to use AI functionalities such as writing the message summary and suggestions on WhatsApp, which are also demanding for on-device hardware.
The new feature will be completely opt-in and will not be able to default, which will give users complete control over how and when they choose it.
Private processing is not available immediately to WhatsApp users, but will gradually be rolled out in the coming weeks.
How does private processing works
For those who choose to use private processing, the system makes an anonymous certification through the user’s WhatsApp client to ensure the validity of the user.
Subsequently, the app receives a public HPKE encryption key from a third-party CDN so that meta specific users do not detect back requests, maintain full oblivion.
The user’s device starts a connection to the meta gateway through the third party relay, hiding its actual IP address. It establishes a remote verification (RA) + TLS session between the user’s device and the reliable execution environment (TEE) of the meta.
Subsequently, the user’s device sends an end-to-end encrypted request for AI data processing using an almanac encryption key, which is processed inside a confidential virtual machine (CVM) apart from meta.
Meta claims that the processing environment is stateless, and all messages are removed after processing, only the “non-sensitive” logs are left behind.
Finally, the AI-reaction is only encrypted with a unique key to the device and processing CVM and is sent back to a safe session for decryption on the user’s device.
WhatsApp is Promise To share CVM binary and some source code to allow external verification, while a wide white paper on the safe design of private processing will also be published soon.
Privacy concerns
Despite the data security and security of the data presented by the meta, there are always concerns when the sensitive data leaves the equipment for processing on the cloud.
Ultimately, unloading AI functions for the cloud server always comes with an underlying risk, even though it is used to apply strong end-to-end encryption.
Users who are uncomfortable with private processing works should disable it.
For those who find advanced AI features useful, but still when the data is allowed to leave their device, it is still needed to remain under control, WhatsApp’s recently launched ‘Advanced Chat Protect’ will be the ideal solution.
Based on the analysis of 14M malicious tasks, search for the top 10 MITERAT & CK techniques behind the 93% attacks and how to defend them against them.
