GHSA-8GMX-W9M8-VX7Q
Contact stored under an own identity's name silently substitutes its keys when the identity is deleted
Texto original en inglés · Ver en español
Ciclo de vida de la vulnerabilidad
| Fecha | Fase | Evento | Detalle |
|---|---|---|---|
| Divulgación | Publicada | GHSA-8GMX-W9M8-VX7Q divulgada | |
| Remediación | Parche 1.4.9 | openssl-encrypt · 1 d desde divulgación | |
| Monitoreo | Detectada | Ingresó al índice | |
| Monitoreo | Verificada | Última verificación contra las fuentes |
Summary
`IdentityStore.get_by_name` resolves `base_path/<name>` (own identities) before `contacts/<name>`, but `add_identity` chose the destination directory purely from `is_own_identity`. Importing a contact whose name matches an existing own identity therefore created a shadowed `contacts/<name>` entry: invisible while the own identity exists, because every lookup resolves the own identity first.
`delete_identity` removed `base_path/<name>` and returned immediately. Deleting the own identity therefore left the shadowed contact in place, and `get_by_name(<name>)` then resolved to the attacker's keys under a name the user trusts. Recipient resolution goes through `get_by_name`, so this is a live key substitution: files encrypted to that name are encrypted to the attacker's key.
`identity list` displayed both entries under the same name with no disambiguation, so the collision was not visible there either.
Impact
An attacker who can get a contact bundle imported under a chosen name obtains silent key substitution at a later, user-initiated moment (the deletion of their own identity — an ordinary action, e.g. after a key rotation). The TOFU key-change dialogue does fire on the import, but it is a gate about a *changed key*, designed to be passable with `--allow-key-change`, and it does not fire at all when the fingerprints happen to match.
The desktop GUI's contact-import alias field (added in the same release series) made this materially easier to reach: instead of the attacker having to guess a name and the user having to accept a warning about a bundle claiming to be their own identity, the user types their own identity's name into a free-text alias box.
Fix
- `add_identity` refuses to store a name that already exists as the other kind, in both directions, independently of the TOFU key-change gate — a name resolves to one key, so the namespace collision fails closed on its own.
- `delete_identity` removes both locations, so a store that already contains a shadow cannot promote it on deletion.
- `IdentityStore.find_shadowed_names()` reports colliding names, and `identity list` surfaces them in both its human output and its `--json` document, so an existing shadow is visible rather than silent.
Mitigation
Upgrade. Note that `identity delete <name>` removes both entries by default — correct for a collision, since leaving a resolvable shadow is the substitution, but it destroys the own identity's private keys. Use `--kind own|contact` to remove one side, and back up the identity store first.
Upgrade. On an affected version, run `identity list` and check for a name appearing as both an own identity and a contact; verify any such contact's fingerprint out of band before deleting anything.
Rangos afectados
Estado del parche por paquete
| Paquete / producto | Estado | Corregida en | Desde divulgación | Última versión |
|---|---|---|---|---|
pypiopenssl-encrypt |
Corregida | 1.4.9 | 1 d | 1.4.9 no afectada |
1.4.9Arsenal público
Sin exploits, módulos ni PoC públicos conocidos a la fecha de verificación.
https://github.com/jahlives/openssl_encrypt/security/advisories/GHSA-8gmx-w9m8-vx7qLas referencias externas se muestran para copiar: cve.llama.ec no lo redirige a otros sitios.
