You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
#3795 added a new method UserIdentity::has_verification_violation.
However, it appears to have omitted support for tracking verification of the user's own identity. This means that if our own identity was rotated, and a new device added, signed by the new identity, we would share messages to that device despite there being no evidence it is not controlled by an attacker.
This will become less of a problem in future, when we refuse to do any operations without our own identity being verified, but in the meantime it is a security hole, and, for as long as matrix-sdk-crypto has support for sending messages without verifying our identity, it will remain a footgun. I think it therefore needs fixing as a prerequisite to #3793.
The text was updated successfully, but these errors were encountered:
This means that if our own identity was rotated, and a new device added, signed by the new identity, we would share messages to that device despite there being no evidence it is not controlled by an attacker.
This is not true with OnlySigned isolation mode, the sending will fail until you verified the new identity:
This is not true with OnlySigned isolation mode, the sending will fail until you verified the new identity:
Yeah. I've slightly forgotten what all this was about, but I think it was mostly about CollectStrategy::DeviceBasedStrategy (ie, the strategy used in AllDevicesIsolationMode):
This will become less of a problem in future, when we refuse to do any operations without our own identity being verified, but in the meantime it is a security hole, and, for as long as matrix-sdk-crypto has support for sending messages without verifying our identity, it will remain a footgun.
#3795 added a new method
UserIdentity::has_verification_violation
.However, it appears to have omitted support for tracking verification of the user's own identity. This means that if our own identity was rotated, and a new device added, signed by the new identity, we would share messages to that device despite there being no evidence it is not controlled by an attacker.
This will become less of a problem in future, when we refuse to do any operations without our own identity being verified, but in the meantime it is a security hole, and, for as long as matrix-sdk-crypto has support for sending messages without verifying our identity, it will remain a footgun. I think it therefore needs fixing as a prerequisite to #3793.
The text was updated successfully, but these errors were encountered: