Open AntoineRondelet opened 4 years ago
If I haven't missed anything, the edward
curve implemented in libff doesn't seem to implement Decaf, and has a cofactor = 4. It may be worth double checking as I believe that the subgroup membership tests may have been forgotten. To double check.
Related to: https://github.com/scipr-lab/libff/issues/45
See also, this recent note by M. Scott: https://eprint.iacr.org/2021/1130.pdf
For now, no "multiplication by the cofactor" is done in the BLS12_377 implementation. However, the cofactor isn't 1:
This is obviously insecure and vulnerable to attacks such as the small subgroup attack. Add multiplication by cofactor or subgroup membership test where relevant - i.e. in the deserialization functions etc - (same applies in G2 and for other curves with cofactor =/= 1)