GHSA-xq78-wvcv-hqvj json
MIKEY message MAC is never verified on receive, so forged messages are accepted
published last mo. modified last mo. also RUSTSEC-2026-0336
- cvss
- 7.5
- version 3.1, OSV
- cisa kev
- Not listed
- not in CISA's catalog
- epss
- —
- no score, EPSS covers CVEs only
- public exploits
- 0
- none known
Summary
`mykey` implements MIKEY (RFC 3830) for SRTP key exchange. No receive path verified the MIKEY message authentication code.
`crypto::verify_mac` existed and was correct, but it was called only from unit tests — neither `MikeyMessage::from_bytes` nor `MikeyMessage::complete_psk` invoked it.
`new_psk_init` computed an HMAC-SHA-1-160 over the serialized message and appended it, so the tag was present on the wire and looked like integrity protection. It was decorative: a receiver accepted any message regardless of whether the MAC was correct, wrong, or absent.
## Consequences
- An attacker able to inject or modify MIKEY messages in transit faced no integrity check. Modified RAND, SSRC or security-policy payloads were accepted silently. Note that this does not require a man-in-the-middle position: since nothing was checked, a wholly forged PSK-Init was accepted. - Pre-shared key mode therefore provided no mutual authentication, despite the documentation claiming that it did. Possession of the pre-shared key was never actually proven by either party. - Replayed init messages were not rejected.
The practical severity was bounded by GHSA-fwjg-cm92-j3hj (cleartext TGK), which already let a passive observer recover the key material outright without needing to forge anything.
## The fix
`complete_psk` now verifies the message MAC **before** any decryption, so a message that fails authentication yields no key material at all. A `MacAlg::Null` payload, an absent MAC and a truncated MAC are each refused rather than treated as "no check required".
The authentication key is now derived from the pre-shared key per RFC 3830 §4.1.4 rather than from the TGK — a prerequisite, since verifying against a TGK-derived key would have been circular while the TGK travelled in the message itself. Optional §6.9 verification messages were also added, giving the exchange genuine mutual authentication.
Replay protection is still absent: a captured PSK-Init can be replayed and will be accepted, because the optional COUNTER timestamp is emitted rather than a mandatory NTP type and no record of seen timestamps or RAND values is kept. That gap is tracked at <https://github.com/waxspin/mykey/issues/24> and is not covered by this advisory.
## Workarounds
Treat parsed MIKEY messages as unauthenticated input: do not rely on pre-shared key mode for peer authentication or message integrity, and carry it over a transport that provides both.
Patch / fix
Is GHSA-xq78-wvcv-hqvj in your dependencies? Fencer shows which of your repositories are affected as soon as an advisory lands, and opens the fix as a pull request.
Check with FencerCVSS scores
| assigned by | version | score | vector |
|---|---|---|---|
| OSV | 3.1 | 7.5 | CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:N |
Affected packages
| ecosystem | package | affected | fixed in |
|---|---|---|---|
| crates.io | mykey | >=0.0.0-0, <2.0.0 |
2.0.0 |
References
Sources
Fencer Intel keeps one entry per vulnerability: the CVE id wins, then GHSA, and records that alias each other are merged, except that two CVEs always keep separate entries. The same data is available as JSON.