Vulnerability CriticalRUSTSEC-2026-0335 · mykey
RUSTSEC-2026-0335: Pre-shared key mode transmits the TGK in cleartext, exposing the SRTP master key
Short answerA security vulnerability in the mykey crate: Pre-shared key mode transmits the TGK in cleartext, exposing the SRTP master key. Critical severity (CVSS 9.1). Fixed in 2.0.0: update to a fixed version.
| Crate | mykey |
|---|---|
| Type | Vulnerability (crypto-failure) |
| Affected versions | Before 2.0.0. |
| Fixed in | 2.0.0 |
| Severity | Critical, CVSS 9.1 CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N |
| Published | 2026-10-09 |
| Last changed | 2026-10-12 |
| Also known as | GHSA-fwjg-cm92-j3hj |
What to do
- See which version you have:
cargo tree -i mykey(it also shows which of your dependencies pulls it in). - Update within your version requirement:
cargo update -p mykey. If the fix is a newer major or minor than your requirement allows, raise it in Cargo.toml (for examplemykey = "2.0.0") or update the dependency that pulls it in. - Check again with
cargo audit, or paste your Cargo.lock into our Cargo.toml checker.
Details from the advisory
mykey implements MIKEY (RFC 3830) for SRTP key exchange. Pre-shared key mode transmitted the TEK Generation Key (TGK) in the clear.
MikeyMessage::new_psk_init built its KEMAC payload with EncAlg::Null and placed the TGK in enc_data verbatim. Anyone able to read a PSK-Init message therefore recovers the TGK , and so the SRTP master key , **without knowing the pre-shared key**.
The transmitted value was never even used by the receiver: complete_psk re-derived the TGK from the pre-shared key and ignored enc_data entirely. The field was write-only; it provided nothing and disclosed everything.
A second consequence: the message authentication key was derived from the TGK rather than from the pre-shared key, so an attacker who read the TGK off the wire could also forge a valid message MAC. Integrity was lost along with confidentiality. That aspect is tracked separately as GHSA-xq78-wvcv-hqvj.
Exploitability
A MIKEY message is normally distributed as an SDP a=key-mgmt:mikey attribute, which in AES67 and SMPTE ST 2110 deployments is typically announced over SAP multicast. A passive listener on the media network recovers the session key with no active attack and no credentials.
RFC 3830 §4.2.3 permits NULL encryption only where "the underlying protocols can guarantee security", and warns to use it with caution. An SDP file or a SAP announcement is not such a channel.
The fix
Pre-shared key mode now follows RFC 3830 §3.1 as key *transport*: the initiator generates a random TGK, frames it as a Key data sub-payload (§6.13), and encrypts it with AES-CM (§4.2.3) under an encryption key and 112-bit salt derived from the pre-shared key (§4.1.4). The message MAC is keyed from the pre-shared key and verified before anything is decrypted.
This changes the wire format, so both ends of an exchange must be upgraded together. The incompatibility is silent: an old peer and a new peer complete the exchange without error and then derive different SRTP keys.
Workarounds
Use ephemeral Diffie-Hellman mode instead of pre-shared key mode, or carry the PSK-Init only over a channel that is already confidential and authenticated end to end.
References
- Package: crates.io/crates/mykey
- Advisory: rustsec.org/advisories/RUSTSEC-2026-0335.html
- Web: github.com/waxspin/mykey/pull/25
From the RustSec Advisory Database (CC0-1.0) via the OSV export, downloaded Oct 24, 2026. Severity score computed from the published CVSS vector.
Other advisories for mykey
Up next
Same cratemykey: security historyEvery advisory for this crate, and whether it is safe to use today.Read it next →Did this answer your question?