Closed vmx closed 1 year ago
This kind of asks for the following question. If the input is not considered trustworthy at this stage, then how come just check for infinity is sufficient? I mean wouldn't it be as appropriate to verify that the corresponding points are in the [pairing-friendly] group? In other words, there seem to be a contradiction, input is either already trusted, in which case no check should be necessary, or it's untrusted, in which case the referred check is arguably insufficient.
I checked with the team. Indeed it doesn't make sense to only have this check => We don't need it on the C++ side of things.
In the C2 code base the
bellperson
patch currently contains a check for a subversion-CRS attachk. This check is currently based on the parameters that are read in on the Rust side.Instead this check should be on the parameters that are actually used for the proof. Hence this check should be moved into this code base.