EC crypto: Equivalent of EC_POINT_invert() in go?

[Anatol Pomozov]

Hi

I am trying to port some of EC crypto code to golang. And I have a call in C

            if (EC_POINT_invert(grp, p, bnc) < 0)

                return NULL;

that I need to port to golang. I checked the crypto lib docs and did not find any mention of EC point inversion operation.

Could anyone point me to the right direction on how to achieve the inversion in Go?

[Joop Kiefte]

https://www.openssl.org/docs/man1.1.0/man3/EC_POINT_invert.html I found this, that might help you on the way...

It's defined in this file: https://github.com/openssl/openssl/blob/67ecd65cc4fdaa03fbae5fcccf53ebca7d785554/crypto/ec/ec_lib.c

Joop Kiefte - Chat @ Spike

--
You received this message because you are subscribed to the Google Groups "golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/golang-nuts/3be55ce9-d6e5-46e7-84cc-3866e7d9b591n%40googlegroups.com.

[Anatol Pomozov]

Hi

On Fri, Sep 25, 2020 at 10:51 AM Joop Kiefte <iko...@gmail.com> wrote:

https://www.openssl.org/docs/man1.1.0/man3/EC_POINT_invert.html I found this, that might help you on the way...

It's defined in this file: https://github.com/openssl/openssl/blob/67ecd65cc4fdaa03fbae5fcccf53ebca7d785554/crypto/ec/ec_lib.c

Or more precisely it calls this code https://boringssl.googlesource.com/boringssl/+/master/crypto/fipsmodule/ec/felem.c#57

But I want to avoid implementing parts of EC encryption myself and looking for a standard Golang way to achieve my goal.

Shouldn't this inversion operation be part of the standard library?

[Conrado P. L. Gouvêa]

On Fri, Sep 25, 2020 at 3:01 PM Anatol Pomozov <anatol....@gmail.com> wrote:
> Or more precisely it calls this code https://boringssl.googlesource.com/boringssl/+/master/crypto/fipsmodule/ec/felem.c#57
>
> But I want to avoid implementing parts of EC encryption myself and looking for a standard Golang way to achieve my goal.
>
> Shouldn't this inversion operation be part of the standard library?
>

My understanding is that the Go standard library prefers that the entire underlying logic of ECC should be hidden from the user, like in "crypto/ed25519".

However, they can't get rid of elliptic.CryptoParams due to compatibility. It's not a good idea in a general sense to use CryptoParams because it uses big.Int which is not side-channel resistant. For these reasons I don't think it's likely they'all add point inversion.

If you do want to go on... you can "invert" a point ("negate" would be a better term...) by simply setting y = n - y, for example:

y := new(big.Int).Sub(curveParams.Params().N, y)

Best regards,

Conrado

[Anatol Pomozov]

Hi

Thank you. I end up with following code for calculating "A - B" expression:

x, y = ecCurve.Add(a.X, a.Y, b.X, new(big.Int).Neg(b.Y))

And it seems to work the same as the openssl version I posted above.
Thank you for your help.

Though having a function in the standard library (either "Sub()" or
"Negate()") would make such code cleaner and easier to follow.