listDevices returns identity-only stubs — enough for a device we already have a session with, useless for one we do not, because initiate() verifies the signed prekey's signature and a stub carries empty bytes. Choosing between the two lists meant that whenever ANY device already had a session, every device got a stub, so a device still needing one could drop out of the fan-out with no partial-delivery warning (the other device succeeded, so was non-empty). The rule is union: full bundle where the key route returned one, stub otherwise. Found while verifying a two-device pair by hand; the live suite's 'a message reaches every device a person has' passes either way, which is why it needed the hand test to surface. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2.1 KiB
Security
Reporting a vulnerability
Email security@flamenetmessenger.com. Please do not open a public issue for anything affecting confidentiality or authentication. We acknowledge within 72 hours.
Status
Not independently audited. This Android app has been reviewed in-house and is covered by the JVM unit suites plus on-device instrumentation, alongside the shared 253-check relay conformance suite and cross-platform interop tests that run the real engine on both sides.
An internal adversarial audit on 2026-08-29 found and fixed a family of sender-attribution flaws: the ratchet authenticates a device, and the user id inside a sealed envelope is a claim its author chooses — the two were not bound. Devices now bind to their user on first proven use, control traffic (delivery keys, avatars) is refused from an unproven sender, and the key-transparency log head is verified against a pinned issuer key. Finding that class of bug in-house is a reason to expect more from outside review, not fewer.
What the app protects, and what it cannot
Private keys are generated on the device and live in the Android Keystore, which never hands out key material. Message history is sealed at rest. The app defends against a malicious or compromised relay reading messages, and against a passive adversary recording traffic to decrypt after a quantum computer exists (when both ends publish a post-quantum identity — the safety number says which guarantee it covers).
It does not defend against:
- A compromised device. Malware with your unlock reads plaintext. The app lock is a speed bump against someone holding your unlocked phone, not encryption.
- Metadata. The relay sees who is addressed and when. Sealed sender removes the sender from the stored row, not from the network connection.
- Your correspondent. Anything you send can be screenshotted, kept, or repeated by the person you sent it to.
Verifying what you run
Every byte of the shipped APK's contents reproduces from this source; only the
randomized signing block differs. See docs/REPRODUCIBLE.md in the relay repo
and scripts/verify-build.py in the Android repo.