Package Instance

matrix-android-sdk2 vulnerable to impersonation via forwarded Megolm sessions
### Impact
An attacker cooperating with a malicious homeserver can construct messages appearing to have come from another person. Such messages will be marked with a grey shield on some platforms, but this may be missing in others.

This attack is possible due to the matrix-android-sdk2 implementing a too permissive [key forwarding](https://spec.matrix.org/v1.3/client-server-api/#key-requests) strategy on the receiving end.

Key forwarding is a mechanism allowing clients to recover from “unable to decrypt” messages when they missed the initial key distribution, at the time the message was originally sent. Examples include accessing message history before they joined the room but also when some network/federation errors have occurred.

### Patches

The default policy for accepting key forwards has been made more strict in the matrix-android-sdk2. The matrix-android-sdk2 will now only accept forwarded keys in response to previously issued requests and only from own, verified devices.

A unique exception to this rule is with the experimental [MSC3061](https://github.com/matrix-org/matrix-spec-proposals/pull/3061), that is forwarding room keys for past messages when invited in a room configured with the proper history visibility setting. Such key forwards are parked upon receipt and are only accepted if the SDK receives an invitation for that room from the inviter in a limited time window. 

The SDK now sets a `trusted` flag on the decrypted message upon decryption, based on whether the key used to decrypt the message was received from a trusted source. Clients need to ensure that messages decrypted with a key with `trusted = false` are decorated appropriately (for example, by showing a warning for such messages).

### Workarounds
Current users of the SDK can disable key forwarding in their forks using `CryptoService#enableKeyGossiping(enable: Boolean)`.

### References
Blog post: https://matrix.org/blog/2022/09/28/upgrade-now-to-address-encryption-vulns-in-matrix-sdks-and-clients

### For more information
If you have any questions or comments about this advisory, e-mail us at [security@matrix.org](mailto:security@matrix.org).

matrix-android-sdk2 vulnerable to Olm/Megolm protocol confusion
### Impact
An attacker cooperating with a malicious homeserver can construct messages that legitimately appear to have come from another person, without any indication such as a grey shield. 

Additionally, a sophisticated attacker cooperating with a malicious homeserver could employ this vulnerability to perform a targeted attack in order to send fake to-device messages appearing to originate from another user. This can allow, for example, to inject the key backup secret during a self-verification, to make a targeted device start using a malicious key backup spoofed by the homeserver. matrix-android-sdk2 would then additionally sign such a key backup with its device key, spilling trust over to other devices trusting the matrix-android-sdk2 device.

These attacks are possible due to a protocol confusion vulnerability that accepts to-device messages encrypted with Megolm instead of Olm.

### Patches

matrix-android-sdk2 has been modified to only accept Olm-encrypted to-device messages and to stop signing backups on a successful decryption.

Out of caution, several other checks have been audited or added:
- Cleartext `m.room_key`, `m.forwarded_room_key` and `m.secret.send` to_device messages are discarded.
- Secrets received from untrusted devices are discarded.
- Key backups are only usable if they have a valid signature from a trusted device (no more local trust, or trust-on-decrypt).
- The origin of a to-device message should only be determined by observing the Olm session which managed to decrypt the message, and not by using claimed sender_key, user_id, or any other fields controllable by the homeserver.


### Workarounds

As this attack requires coordination between a malicious home server and an attacker, if you trust your home server no particular workaround is needed.  Notice that the backup spoofing attack is a particularly sophisticated targeted attack.

We are not aware of this attack being used in the wild, though specifying a false positive-free way of noticing malicious key backups key is challenging.

As an abundance of caution, to avoid malicious backup attacks, you should not verify your new logins using emoji/QR verifications methods until patched. Prefer using verify with passphrase.


### References
Blog post: https://matrix.org/blog/2022/09/28/upgrade-now-to-address-encryption-vulns-in-matrix-sdks-and-clients

### For more information
If you have any questions or comments about this advisory, e-mail us at [security@matrix.org](mailto:security@matrix.org).