4 ms·

I don't know the implementation on the client, but in theory it can be implemented locally on the client only, and I made the above statements under the assumpt

by est31 6y ago

I don't know the implementation on the client, but in theory it can be implemented locally on the client only, and I made the above statements under the assumption that it's how it's implemented. The user issues the report command in the UI and the client takes its local copy and uploads it to the server as part of the report. So it's mainly a convenience feature. The only case where the current E2EE security model is changed is when the protocol gains a feature to enable the client to attach a cryptographic proof about the message's content. That would violate the non-repudiation properties of the Signal protocol, but it would make reports more believable as you can't fake them any more with a modified client. As I've said I don't know the implementation, but it's entirely possible that they didn't include any cryptographic proof and keep it equivalent to screenshotting.

upofadown 6y ago
>That would violate the non-repudiation properties of the Signal protocol,... Signal achieves non-repudiation by allowing forgery of messages after they have been transferred and verified by the receiver. It is not possible to undo that. You would have to make a special exception at the time the message was transferred which wouldn't be practical in this case. This all assumes that you believe that establishing forgeability actually achieves non-repudiation in the first place. Facebook might want to pass along the signature information anyway.
varispeed 6y ago
So is there a way to capture user's screen without user knowing and essentially bypassing the end to end encryption altogether? (the images would be sent to a different endpoint for example)
- est31 6y ago
  The report feature is initiated by the user. Only the copy of the user's correspondence who sends the report is sent. If there is a risk that the user doesn't know that the correspondence is sent if they hit the report button, then fb should add a warning, but the feature itself is fine IMO.

by est31 6y ago

upofadown 6y ago

>That would violate the non-repudiation properties of the Signal protocol,... Signal achieves non-repudiation by allowing forgery of messages after they have been transferred and verified by the receiver. It is not possible to undo that. You would have to make a special exception at the time the message was transferred which wouldn't be practical in this case. This all assumes that you believe that establishing forgeability actually achieves non-repudiation in the first place. Facebook might want to pass along the signature information anyway.

varispeed 6y ago

So is there a way to capture user's screen without user knowing and essentially bypassing the end to end encryption altogether? (the images would be sent to a different endpoint for example)

est31 6y ago
The report feature is initiated by the user. Only the copy of the user's correspondence who sends the report is sent. If there is a risk that the user doesn't know that the correspondence is sent if they hit the report button, then fb should add a warning, but the feature itself is fine IMO.