Could be of interest to you:
https://github.com/bitcoin-core/secp256k1/pull/767
Not applicable to this thread. It's about fault attacks, where you cause a glitch in a cpu while signing to make it miscompute and leak a key. They demonstrate a particular form of this against several curves.
I don't think their countermeasure advice is all that great. Sure, storing G as x/y is cheap and stops this particular approach (except in ecdh) so everyone should do that and effectively everything does, but if an attacker can cause skipped instructions there probably are a bunch of other ways to attack. Better to verify after signing, like bitcoin core does.
operation at line 5 typically fails in that case (either because
the square root algorithm fails on nonquadratic residues, or
because the resulting point fails point validation). This implies
that, for example, secp192k1 and secp256k1 are susceptible to
the SCPD attack, but secp224k1 is not.
EDIT wtere are many different attack for ex chnge curve, change random euation of ecps256k1 to linear form, use more faster then ecps256k1.lib math operation realisation etc.
Q. Did someone know how to modify public key(for ex "split publick key") for get smaler byte range of privkey