Labyrinth 1.1：エンドツーエンド暗号化バックアップの信頼性をさらに向上Labyrinth 1.1: Making End-to-End Encrypted Backups Even More Reliable

Meta has started rolling out version 1.1 of Labyrinth, the encrypted storage system and protocol that secures message contents and conversation history on Messenger. The update focuses on improving the reliability and fault tolerance of end-to-end encrypted (E2EE) backups — an unglamorous but essential layer of the messaging stack, since a backup that fails to restore when a device is lost or replaced effectively negates the value of keeping an encrypted history at all.

Labyrinth serves as the foundation for storing Messenger history so that it remains readable only by the account owner, even though the encrypted data resides on Meta's servers. When Meta moved to make E2EE the default for personal chats and calls on Messenger, it had to solve a problem that secure messengers have long faced: how to let people recover their conversations on a new phone without giving the provider the ability to read them. Storing message history in plaintext on the server is simple but breaks the encryption guarantee, while keeping it only on the device risks permanent loss. Labyrinth is Meta's answer to that tension.

In broad terms, E2EE backup systems work by encrypting data on the device with a key the server never sees. That key is typically protected by a user-chosen secret such as a PIN, or by a longer randomly generated recovery code that the user is expected to store safely. The encrypted material is uploaded to the provider's infrastructure, and access is gated so the secret cannot simply be guessed. Systems of this kind commonly rely on hardware security modules or secure enclaves to enforce strict limits on the number of decryption attempts, which prevents an attacker — or the provider itself — from brute-forcing a short PIN.

The 1.1 release appears to concentrate on operational robustness rather than changes to the cryptographic guarantees themselves. Improvements to reliability and fault tolerance in a system like this likely involve better handling of partial failures during upload or restore, redundancy so that a single component outage does not block recovery, and more resilient synchronization when a user has multiple devices or restores after a long gap. For E2EE backups, these engineering details matter because the provider cannot inspect or repair the contents directly; the encryption that protects users also constrains what operators can do to recover from errors, so correctness has to be built in from the start.

The work fits into a broader industry pattern of extending end-to-end encryption from messages in transit to data at rest. WhatsApp, also owned by Meta, introduced encrypted chat backups in 2021, protecting them with either a password or a 64-digit key and backing the system with a key vault built on hardware security modules. Apple's Advanced Data Protection, launched in 2022, similarly extends end-to-end encryption to iCloud backups and other categories of data, with recovery tied to user-held keys or trusted contacts. Signal has long used a PIN-based secure value recovery scheme, originally built on secure enclaves, to let users restore certain account data. Labyrinth places Messenger within this same design space.

These approaches share a common trade-off that is worth understanding. Stronger encryption shifts responsibility toward the user: if a PIN or recovery code is lost and no fallback is available, the data may be unrecoverable by design, because the provider holds no key. That is the point of the architecture, but it raises the stakes for the surrounding recovery and reliability machinery — exactly the area Labyrinth 1.1 is said to target. Making restores succeed consistently is, in practice, what determines whether users keep encrypted backups enabled at all.

Meta is publishing the update through its engineering blog, consistent with how it has documented other security and infrastruct