Skip to content

Latest commit

 

History

History
133 lines (107 loc) · 6.65 KB

causal-03-freshness.rst

File metadata and controls

133 lines (107 loc) · 6.65 KB

Freshness and presence

Non-causal drops

As mentioned at the end of the previous section, we are so far unable to detect non-causal drops - drops of messages that are not an ancestor of messages we have seen and delivered. This does not affect transcript consistency, as already argued. However, sometimes we might still want this as a security property, especially in low-latency protocols.

One way we can do this is via fresh nacks - messages that we can be sure are fresh (as of time t), that say "no messages sent by u since t". Our by(u) total ordering encodes the "no messages since" meaning implicitly, so we just need to prove freshness since time t.

A message m implies its sender was present at time of context(m)(self), if it exists - we trust our own timestamps, and the partial order establishes the "after" relationship. If context(m)(self) does not exist, then we'll have to use something else written by ourselves, such as e.g. a key exchange message that helped to derive the secrets that m is auth-encrypted by - we don't have a mechanism to securely trust others' timestamps, and won't try to construct one here. Note that we just store our timestamps locally, we don't need to send them anywhere - and they wouldn't trust ours anyway.

The above achieves end-to-end security for freshness. By contrast, XMPP presence is not end-to-end secure - it assumes a trusted server. Neither is looking at "local receive time of a packet" - it assumes a trusted network.

Heartbeats

We detect non-causal drops by periodically sending out heartbeats, and waiting for these to be fully-acked (i.e. setting an ack-monitor on them); this proves freshness as of the time of the heartbeat. To be efficient, we can omit a heartbeat if we've already sent another message inside the previous period, excluding explicit acks which don't need to be acked. We may refer to these messages as "implicit heartbeats", akin to implicit acks.

If someone takes too long to ack a (implicit or explicit) heartbeat, we can indicate this in the UI, for example by dulling the user avatar.

If heartbeats are regular, they may be distinguished from actual messages even to an eavesdropper who can't read their contents. (The same may be said of regular explicit acks and identical-ciphertext resends, though these probably reveal less information.) In general, metadata security is a more complex topic, and it is unclear whether these things are a real security problem to be worth spending effort on, so we'll note the issue but won't pursue it here.

Presence and expiry

Freshness is an indicator of presence, which we'll define roughly as "a good chance a user will ack messages sent to them". We adopt end-to-end definitions, so "presence" is considered from the local user's perspective, rather than the internet - if a user's connection goes down, from their POV everyone else is absent, rather than themselves. Typically, presence statuses must expire unless explicitly refreshed, to be able to reliably indicate absence.

Our heartbeats are a rudimentary indicator of presence, one that is only partly useful since it works within an already-established session. There are many end-to-end secure presence mechanisms that are more general purpose, e.g. TODO. Yet other mechanisms are simple but still effective - e.g. detecting that the local network interfaces are disconnected, is a strong indicator of absence.

Locally, we may integrate our heartbeats with these other mechanisms, for more accuracy and reliability. The exact logic will depend on the mechanism and what interfaces it provides for integration, but it should be fairly intuitive and straightforward. For example, the combined system should indicate presence when a user acks a heartbeat or any external mechanism indicates presence, and any of these should reset any expiry-based absence indicators.

Absence is semantically distinct from parting the session - absent users still have the right to see messages, until they formally stop being a member of the session. This distinction supports the ability to have a long-running session where users go online and offline, but still want to see what was said during their absence. This is common for some existing hosted chat systems, but we can do this end-to-end too.

The feature of long-term absences must be considered against the requirements of the greeting protocol. For example, some GKEs require all members to be present in order to complete execution. Then, supporting long-term absences is problematic, because this support will likely result in most of the session's lifetime having partial membership presence. To avoid this, we recommended that systems choose a greeting protocol that supports the forced-parting of absent members without their input.

If a session is active whilst a user is absent, this will cause non-full-ack warnings to be emitted. But this is expected because they are absent, so we can account for it: if the users who haven't acked a message are all absent, the severity of its non-full-ack warning should be reduced. Furthermore, the rate of resends should be reduced too, since we believe they won't succeed anyway. (However, if we have no other external presence mechanism, they should not be completely stopped - otherwise all parts of a partition will stop sending, and we'll never regain presence.) Later, when the absent user returns, the rate of resends should be restored, or even reset to the maximum.

Timestamps

So far we have avoided communicating any timestamps in the protocol, nor to rely on specific timestamp values for others' messages. (Heartbeat timestamps are only used locally as a lower-bound.)

However, specific values for timestamps are useful as a UI indication. How do we do this?

Limitations

TODO Clock reliability.

  • new device, time not set
  • when travelling, some users change system clock instead of time zone
  • change system clock for testing and forget about it

Verifiability.

If a member lies in the contents of a message, this is not something the messaging protocol can detect. We treat timestamps as a similar sort data - no other part of the system or the security models depends on them being even vaguely correct.

Simplest option is to use local delivery time of a message. Additionally, each message may contain the claimed sender local time, which may be displayed on the side.

See appendix (TODO: link) for discussion of some more complex techniques, which may result in better accuracy when users have skewed clocks. However, they are quite complex and don't achieve very strong end-to-end security (as mentioned above), which is why they are only in the appendix.