Naming elliptic curves used in cryptography

There are an infinite number of elliptic curves, but a small number that are used in elliptic curve cryptography (ECC), and these special curves have names. Apparently there are no hard and fast rules for how the names are chosen, but there are patterns.

The named elliptic curves are over a prime field, i.e. a finite field with a prime number of elements p, denoted GF(p). The number of points on the elliptic curve is on the order of p [1].

The curve names usually contain a number which is the number of bits in the binary representation of p. Let’s see how that plays out with a few named elliptic curves.

Curve nameBits in p
ANSSI FRP256v1  256
BN(2, 254)254
brainpoolP256t1  256
NIST P-224224

In Curve25519, p = 2255 – 19 and in Curve 383187, p = 2383 – 187. Here the number of bits in p is part of the name but another number is stuck on.

The only mystery on the list is Curve1174 where p has 251 bits. The equation for the curve is

x² + y² = 1 – 1174 y²

and so the 1174 in the name comes from a coefficient rather than from the number of bits in p.

Edwards curves

The equation for Curve1174 doesn’t look like an elliptic curve. It doesn’t have the familiar (Weierstrass) form

y² = x³ + ax + b

It is an example of an Edwards curve, named after Harold Edwards. So are all the curves above whose names start with “E”. These curves have the form

x² + y² = 1 + d x² y².

where d is not 0 or 1. So some Edwards curves are named after their d parameter and some are named after the number of bits in p.

It’s not obvious that an Edwards curve can be changed into Weierstrass form, but apparently it’s possible; this paper goes into the details.

The advantage of Edwards curves is that the elliptic curve group addition has a simple, convenient form. Also, when d is not a square in the underlying field, there are no exceptional points to consider for group addition.

Is d = -1174 a square in the field underlying Curve1174? For that curve p = 2251 – 9, and we can use the Jacobi symbol code from earlier this week to show that d is not a square.

    p = 2**251 - 9
    d = p-1174
    print(jacobi(d, p))

This prints -1, indicating that d is not a square. Note that we set d to p – 1174 rather than -1174 because our code assumes the first argument is positive, and -1174 and p – 1174 are equivalent mod p.

Update: More on addition on Curve1174.

Prefix conventions

A US government publication (FIPS PUB 186-4) mandates the following prefixes:

  • P for curves over a prime field,
  • B for curves over a binary field (i.e. GF(2n)), and
  • K for Koblitz fields.

The ‘k’ in secp256k1 also stands for Koblitz.

The M prefix above stands for Montgomery.

Related posts

[1] It is difficult to compute the exact number of points on an elliptic curve over a prime field. However, the number is roughly p ± 2√p. More precisely, Hasse’s theorem says

|\#(E/\mathbb{F}_p) - p - 1| \leq 2\sqrt{p}

One thought on “Naming elliptic curves used in cryptography

Leave a Reply

Your email address will not be published. Required fields are marked *