MSC4147: Including device keys with Olm-encrypted events (#4147)

proposals/4147-including-device-keys-with-olm-encrypted-events.md

@@ -0,0 +1,183 @@
+# MSC4147: Including device keys with Olm-encrypted to-device messages
+
+Summary: a proposal to ensure that messages sent from a short-lived (but
+genuine) device can be securely distinguished from those sent from a spoofed
+device.
+
+## Background
+
+When a Matrix client receives an encrypted message, it is necessary to
+establish whether that message was sent from a device genuinely belonging to
+the apparent sender, or from a spoofed device (for example, a device created by
+an attacker with access to the sender's account such as a malicious server
+admin, or a man-in-the-middle).
+
+In short, this is done by requiring a signature on the sending device's device
+keys from the sending user's [self-signing cross-signing
+key](https://spec.matrix.org/v1.12/client-server-api/#cross-signing). Such a
+signature proves that the sending device was genuine.
+
+Current client implementations check for such a signature by
+[querying](https://spec.matrix.org/v1.12/client-server-api/#post_matrixclientv3keysquery)
+the sender's device keys when an encrypted message is received.
+
+However, this does not work if the sending device logged out in the time
+between sending the message and it being received. This is particularly likely
+if the recipient is offline for a long time. In such a case, the sending server
+will have forgotten the sending device (and any cross-signing signatures) by
+the time the recipient queries for it. This makes the received message
+indistinguishable from one sent from a spoofed device.
+
+Current implementations work around this by displaying a warning such as "sent
+by a deleted or unknown device" against the received message, but such
+messaging is unsatisfactory: a message should be either trusted or not.
+
+We propose to solve this by including a copy of the device keys in the
+Olm-encrypted message, along with the cross-signing signatures, so that the
+recipient does not have to try to query the sender's keys.
+
+## Proposal
+
+The plaintext payload of to-device messages encrypted with the [`m.olm.v1.curve25519-aes-sha2` encryption
+algorithm](https://spec.matrix.org/v1.12/client-server-api/#molmv1curve25519-aes-sha2)
+is currently of the form:
+
+```json
+{
+  "type": "<type of the plaintext event>",
+  "content": "<content for the plaintext event>",
+  "sender": "<sender_user_id>",
+  "recipient": "<recipient_user_id>",
+  "recipient_keys": {
+    "ed25519": "<our_ed25519_key>"
+  },
+  "keys": {
+    "ed25519": "<sender_ed25519_key>"
+  }
+}
+```
+
+We propose to add a new property: `sender_device_keys`, which is a copy of what
+the server would return in response to a
+[`/keys/query`](https://spec.matrix.org/v1.12/client-server-api/#post_matrixclientv3keysquery)
+request, as the device keys for the sender's device.  In other words, the
+plaintext payload will now look something like:
+
+```json
+{
+  "type": "<type of the plaintext event>",
+  "content": "<content for the plaintext event>",
+  "sender": "<sender_user_id>",
+  "recipient": "<recipient_user_id>",
+  "recipient_keys": {
+    "ed25519": "<our_ed25519_key>"
+  },
+  "keys": {
+    "ed25519": "<sender_ed25519_key>"
+  },
+  "sender_device_keys": {
+    "algorithms": ["<supported>", "<algorithms>"],
+    "user_id": "<user_id>",
+    "device_id": "<device_id>",
+    "keys": {
+      "ed25519:<device_id>": "<sender_ed25519_key>",
+      "curve25519:<device_id>": "<sender_curve25519_key>"
+    },
+    "signatures": {
+      "<user_id>": {
+        "ed25519:<device_id>": "<device_signature>",
+        "ed25519:<ssk_id>": "<ssk_signature>",
+      }
+    }
+  }
+}
+```
+
+If this property is present, the `keys`.`ed25519` property of the plaintext
+payload must be the same as the `sender_device_keys`.`keys`.`ed25519:<DEVICEID>`
+property.  If they differ, the recipient should discard the event.
+
+As the `keys` property is now redundant, it may be removed in a future version
+of the Matrix specification.
+
+## Potential issues
+
+Adding this property will increase the size of the to-device message.  We found it
+increased the length of a typical encrypted `m.room_key` message from about 1400 to 2400
+bytes (a 70% increase). This will require increased storage on the recipient
+homeserver, and increase bandwidth for both senders and recipients. See
+[Alternatives](#alternatives) for discussion of mitigation strategies.
+
+This proposal is not a complete solution. In particular, if the sender resets
+their cross-signing keys, and also logs out the sending device, the recipient
+still has no way to verify the sending device. The device signature in the Olm
+message is meaningless. A full solution would require the recipient to be able
+to obtain a history of cross-signing key changes, and to expose that
+information to the user; that is left for the future.
+
+## Alternatives
+
+### Minor variations
+
+The `sender_device_keys` property could be added to the cleartext.  That is, it could
+be added as a property to the `m.room.encrypted` event.  This information is
+already public, as it is accessible from `/keys/query` (while the device is
+logged in), and does not need to be authenticated as it is protected by the
+self-signing signature, so it does not seem to need to be encrypted. However,
+there seems to be little reason not to encrypt the information. In addition, by
+including it in the encrypted payload, it leaves open the possibility of
+it replacing the `keys` property, which must be part of the encrypted payload
+to prevent an [unknown key-share attack](https://github.com/element-hq/element-web/issues/2215).
+
+The `sender_device_keys` property could be added to the cleartext by the sender's
+homeserver, rather than by the sending client. Possibly within an `unsigned`
+property, as that is where properties added by homeservers are customarily
+added.  It is not clear what advantage there would be to having this
+information being added by the client.
+
+To mitigate the increased size of to-device events under this proposal, the
+`sender_device_keys` could be sent only in pre-key messages (Olm messages
+with `type: 0` in the `m.room.encrypted` event) — with the rationale that if
+the Olm message is a normal (non-pre-key) message, this means that the
+recipient has already decrypted a pre-key message that contains the
+information, and so does not need to be re-sent the information), or if the
+signatures change (for example, if the sender resets their cross-signing keys),
+or if the sender has not yet sent their `device_keys`.  However, this requires
+additional bookkeeping, and it is not clear whether this extra complexity is
+worth the reduction in bandwidth.
+
+### Alternative approach
+
+A more radical proposal to decrease the overhead in to-device messages is to
+instead specify that `/keys/query` must include deleted devices as well as
+active ones, so that they can be reliably queried. Since the origin server
+might be unreachable at the time the recipient receives the message, such
+device lists would need to be cached on the recipient homeserver.
+
+In other words, this approach would require all homeservers to keep a permanent
+record of all devices observed anywhere in the federation, at least for as long
+as there are undelivered to-device events from such devices.
+
+Transparently: we have not significantly explored this approach. We have a
+working solution, and it is unclear that the advantages of this alternative
+approach outweigh the opportunity cost and delay in rollout of an important
+security feature. If, in future, the overhead of including the device keys
+in the to-device messages is found to be significant, it would be worth
+revisiting this.
+
+## Security considerations
+
+If a device is logged out, there is no indication why it was logged out.  For
+example, an attacker could steal a device and use it send a message.  The user,
+upon realizing that the device has been stolen, could log out the device, but
+the message may still be sent, if the user does not notice the message and
+redact it.
+
+## Unstable prefix
+
+Until this MSC is accepted, the new property should be named
+`org.matrix.msc4147.device_keys`.
+
+## Dependencies
+
+None