Since V2 a contact is its key, there is no address-to-key binding left to verify.
The JSONRPC and FFI APIs are unchanged and report nothing as verified.
BREAKING CHANGE: the statistics JSON sent to the self-reporting-bot on Android changes: Contacts have `encrypted` instead of `verified` and lose `transitive_chain` properties and message stats have `encrypted` instead of `verified` and `unverified_encrypted`, and securejoin invites lose `already_verified`. The collecting bot stores incoming reports verbatim but analysis will have to make sense of older and newer reports.
Headers like From and Autocrypt are now added late,
right before sending the message over SMTP.
This way we advertise the latest list of transports
and use the correct From address in the encrypted part
even for messages queued while being offline.
BCC-self recipients are also added late.
For unencrypted messages we only want to send a copy
to the sending address, but we don't know the sending address
when queueing the message.
Adding bcc-self recipients when dequeuing the message
also makes it possible to send copies to updated list of relays.
According to
<https://datatracker.ietf.org/doc/html/rfc8098#section-3.2.3>
Original-Recipient field values in MDNs MUST NOT be included
if the information about original recipient is not available.
Original recipient may be obtained from ORCPT parameter
of SMTP envelope or from Original-Recipient header
which MTAs are expected to convert Original-Recipient to.
The way we have been using Original-Recipient field is not correct.
Technically we should look for Original-Recipient header
on the message when downloading it from IMAP
and then copy the value into MDN,
but simply assuming it is never there is more correct
than always assuming it is the same as our current address.
I also grepped for Original-Recipient
and removed it together with Reporting-UA from the tests.
Orignal-Recipient is now only left in NDN (bounce messages) test data,
there it is correct as this field is added by MTAs
that have direct access to ORCPT parameter.
Unencrypted messages are no longer sent as multipart/mixed
with Message-ID header duplicated inside.
This was useful as non-standard protection
against servers that rewrite the Message-ID,
but now most messages are sent encrypted
and for them standard RFC 9788 header protection
works the same way.
This value is not used for any logic
except the disappearing messages timer rollback prevention.
It is unlikely that non-chat clients
participate in chats with disappearing messages,
and the case was not tested anyway.
Adds `assert_warn`, `assert_error` and `assert_warns_or_errors`
methods to `TestContext`, that let us assert that
a certain warning or error is logged during the test.
Also asserts test logs should not contain any other
errors or warnings.
Adjusts tests accordingly.
Signed-off-by: Jagoda Ślązak <jslazak@jslazak.com>
- treat an Autocrypt header as a kind of Autocrypt-Gossip: addr=<self>
letting it participate in the existing cooperative key gossipping mechanics.
- speed up processing of incoming autocrypt-gossip headers by using
1 sql commit for N gossip headers instead of the current N commits.
Unifies internal naming of single/direct/1:1/normal chats
to match the naming used in the API.
Closes: #8368
Signed-off-by: Jagoda Ślązak <jslazak@jslazak.com>
---------
Signed-off-by: Jagoda Ślązak <jslazak@jslazak.com>
The `FolderMeaning` and the `target_folder()` function and related code
were needed for the message moving logic. The message moving logic isn't
needed anymore, so, this PR is a first step to remove it.
We only ever look at one folder by now (usually the Inbox folder), so,
any message moving logic isn't needed anymore because we will never
notice a message that is supposed to be moved.
The `target` column of the `imap` table is mostly the same as the
`folder` column now. Except when the message is supposed to be deleted,
then it's an empty string.
Arbitrary timeouts often result in flaky tests,
especially if CI runners may be paused.
In this case however timeout was not used by any tests
and only slowed down the tests in cases where
no message is expected but non-zero timeout is set.
Fix https://github.com/chatmail/core/issues/8195
Supersedes https://github.com/chatmail/core/pull/8217
The most interesting change is in `delete_expired_imap_messages()`
because there, messages are marked for deletion on single-device
chatmail profiles.
The logic in `delete_expired_imap_messages()` was wrong before, and
pre-messages were not deleted at all. This is fixed by also querying
`pre_rfc724_mid` in addition to `rfc724_mid` from the `msgs` table.
In order not to make the SQL statement too complex, I split it into two.
WRT tests:
- `test_immediate_autodelete()` tests the auto-deletion;
`test_imap_autodelete_fully_downloaded_msg()` tests that even for a
message that is split into pre- and post-message, both messages are
deleted.
- A lot of tests test that if bcc_self is on, messages are not
auto-deleted. E.g. `test_one_account_send_bcc_setting()` and
`test_markseen_message_and_mdn()` relies on the fact that messages stay
on the server when bcc_self is on.
- Unfortunately, it is not possible to test that messages are only
deleted for chatmail servers, because we don't have any tests that use a
non-chatmail server.
This change is a preparation for ignoring
unencrypted messages by default.
New test_utils::encrypt_raw_message and
test_utils::receive_encrypted_imf are
used to encrypt the messages before
"receiving" them with receive_imf.
This makes such contacts appear in the contact list. `IncomingTo` is used because
`ChatId::accept_ex()` does so for groups, so sending a group message is effectively accepting the
group again in regards to contact searchability.
This fixes up b549e7633d which made it impossible to find contacts
from groups even if we've written there.
The part of logic there adjusting the sort timestamp forward if the parent message has a greater
sort timestamp wasn't tested explicitly by any test. I only saw one unrelated "golden test" failure
when commented it out.
(Related to #8027)
We know which message was added from the return value
of receive_imf(). It may be that the first chat
in the chatlist is not the one where the message was received
if there is a pinned chat or if
just received message is old.
For now, do this only for `OneOneChat` and `MailingListOrBroadcast`, this is enough to correctly
support messages from modern Delta Chat versions sending Intended Recipient Fingerprint subpackets
and single-recipient messages from modern versions of other MUAs.
This is an important thing forgotten to be checked in 332527089.
Also there's another test which currently doesn't work as we want: outgoing encrypted messages
continue to arrive to ad-hoc group even if we already have contact's key. This should be fixed by
sending and receiving Intended Recipient Fingerprint subpackets.
The good thing is that apparently there are no scenarios requiring the contact to update their
software, the user should just update own devices.
I.e. create a non-replyable ad-hoc group in such cases. Unencrypted replies to encrypted messages
are a security issue and "Composing a Reply Message" from RFC 9787 and "Replying and Forwarding
Guidance" from RFC 9788 forbid such replies.
84161f4202 promotes group members to `Origin::IncomingTo` when
accepting it, instead of `CreateChat` as before, but this changes almost nothing because it happens
rarely that the user only accepts a group and writes nothing there soon. Now if a message has
multiple recipients, i.e. it's a 3-or-more-member group, or if it's a broadcast message, we don't
scale up its recipients to `Origin::OutgoingTo`.
When accepting a chat, its members are promoted to `Origin::CreateChat`, but for groups it makes
sense to use lower origin because users don't always check all members before accepting a chat and
may not want to have the group members mixed with existing contacts. `IncomingTo` fits here by its
definition: "additional To:'s of incoming message of known sender", i.e. we assume that the sender
of some message is known to the user. This way we can show contacts coming from groups in the bottom
of contact list, maybe even add some separator later. It makes sense not to hide such contacts
completely, otherwise if the user remembers the contact name, but not the chat it's a member of, it
would be difficult to find the contact.
This records the curent behavior: after accepting a group with unknown contacts the contact list
contains them mixed with contacts for which 1:1 chats exist, i.e. new contacts from the group are
neither hidden nor sorted down.
Before, outgoing self-sent unencrypted messages were assigned to the self-chat. Now we assign them
to ad-hoc groups with only SELF instead of 1:1 chats with address contacts corresponding to our own
addresses because we don't want to create such address contacts; we still use SELF for `from_id` of
such messages. Not assigning such messages to the encrypted chat should be safe enough and such
messages can actually be sent by the user from another MUA.
It must be verified by "unknown verifier" instead. But if the verifier has known verifier in turn,
it must reverify contacts having unknown verifier. Add a check for this also.
Follow-up for https://github.com/chatmail/core/pull/7042, part of
https://github.com/chatmail/core/issues/6884.
This will make it possible to create invite-QR codes for broadcast
channels, and make them symmetrically end-to-end encrypted.
- [x] Go through all the changes in #7042, and check which ones I still
need, and revert all other changes
- [x] Use the classical Securejoin protocol, rather than the new 2-step
protocol
- [x] Make the Rust tests pass
- [x] Make the Python tests pass
- [x] Fix TODOs in the code
- [x] Test it, and fix any bugs I find
- [x] I found a bug when exporting all profiles at once fails sometimes,
though this bug is unrelated to channels:
https://github.com/chatmail/core/issues/7281
- [x] Do a self-review (i.e. read all changes, and check if I see some
things that should be changed)
- [x] Have this PR reviewed and merged
- [ ] Open an issue for "TODO: There is a known bug in the securejoin
protocol"
- [ ] Create an issue that outlines how we can improve the Securejoin
protocol in the future (I don't have the time to do this right now, but
want to do it sometime in winter)
- [ ] Write a guide for UIs how to adapt to the changes (see
https://github.com/deltachat/deltachat-android/pull/3886)
## Backwards compatibility
This is not very backwards compatible:
- Trying to join a symmetrically-encrypted broadcast channel with an old
device will fail
- If you joined a symmetrically-encrypted broadcast channel with one
device, and use an old core on the other device, then the other device
will show a mostly empty chat (except for two device messages)
- If you created a broadcast channel in the past, then you will get an
error message when trying to send into the channel:
> The up to now "experimental channels feature" is about to become an officially supported one. By that, privacy will be improved, it will become faster, and less traffic will be consumed.
>
> As we do not guarantee feature-stability for such experiments, this means, that you will need to create the channel again.
>
> Here is what to do:
> • Create a new channel
> • Tap on the channel name
> • Tap on "QR Invite Code"
> • Have all recipients scan the QR code, or send them the link
>
> If you have any questions, please send an email to delta@merlinux.eu or ask at https://support.delta.chat/.
## The symmetric encryption
Symmetric encryption uses a shared secret. Currently, we use AES128 for
encryption everywhere in Delta Chat, so, this is what I'm using for
broadcast channels (though it wouldn't be hard to switch to AES256).
The secret shared between all members of a broadcast channel has 258
bits of entropy (see `fn create_broadcast_shared_secret` in the code).
Since the shared secrets have more entropy than the AES session keys,
it's not necessary to have a hard-to-compute string2key algorithm, so,
I'm using the string2key algorithm `salted`. This is fast enough that
Delta Chat can just try out all known shared secrets. [^1] In order to
prevent DOS attacks, Delta Chat will not attempt to decrypt with a
string2key algorithm other than `salted` [^2].
## The "Securejoin" protocol that adds members to the channel after they
scanned a QR code
This PR uses the classical securejoin protocol, the same that is also
used for group and 1:1 invitations.
The messages sent back and forth are called `vg-request`,
`vg-auth-required`, `vg-request-with-auth`, and `vg-member-added`. I
considered using the `vc-` prefix, because from a protocol-POV, the
distinction between `vc-` and `vg-` isn't important (as @link2xt pointed
out in an in-person discussion), but
1. it would be weird if groups used `vg-` while broadcasts and 1:1 chats
used `vc-`,
2. we don't have a `vc-member-added` message yet, so, this would mean
one more different kind of message
3. we anyways want to switch to a new securejoin protocol soon, which
will be a backwards incompatible change with a transition phase. When we
do this change, we can make everything `vc-`.
[^1]: In a symmetrically encrypted message, it's not visible which
secret was used to encrypt without trying out all secrets. If this does
turn out to be too slow in the future, then we can remember which secret
was used more recently, and and try the most recent secret first. If
this is still too slow, then we can assign a short, non-unique (~2
characters) id to every shared secret, and send it in cleartext. The
receiving Delta Chat will then only try out shared secrets with this id.
Of course, this would leak a little bit of metadata in cleartext, so, I
would like to avoid it.
[^2]: A DOS attacker could send a message with a lot of encrypted
session keys, all of which use a very hard-to-compute string2key
algorithm. Delta Chat would then try to decrypt all of the encrypted
session keys with all of the known shared secrets. In order to prevent
this, as I said, Delta Chat will not attempt to decrypt with a
string2key algorithm other than `salted`
BREAKING CHANGE: A new QR type AskJoinBroadcast; cloning a broadcast
channel is no longer possible; manually adding a member to a broadcast
channel is no longer possible (only by having them scan a QR code)
We already have both rand 0.8 and rand 0.9
in our dependency tree.
We still need rand 0.8 because
public APIs of rPGP 0.17.0 and Iroh 0.35.0
use rand 0.8 types in public APIs,
so it is imported as rand_old.
We do not try to delete resent messages
anymore. Previously resent messages
were distinguised by having duplicate Message-ID,
but future Date, but now we need to download
the message before we even see the Date.
We now move the message to the destination folder
but do not fetch it.
It may not be a good idea to delete
the duplicate in multi-device setups anyway,
because the device which has a message
may delete the duplicate of a message
the other device missed.
To avoid triggering IMAP busy move loop
described in the comments
we now only move the messages
from INBOX and Spam folders.
