/fastecdsa

Python library for fast elliptic curve crypto

Primary LanguagePythonThe UnlicenseUnlicense

fastecdsa

https://travis-ci.org/AntonKueltz/fastecdsa.svg?branch=master

This is a python package for doing fast elliptic curve cryptography, specifically digital signatures.

I am not aware of any current issues. There is no nonce reuse, no branching on secret material, and all points are validated before any operations are performed on them. Timing side challenges are mitigated via Montgomery point multiplication. Nonces are generated per RFC6979. The default curve used throughout the package is P256 which provides 128 bits of security. If you require a higher level of security you can specify the curve parameter in a method to use a curve over a bigger field e.g. P384. All that being said, crypto is tricky and I'm not beyond making mistakes. Please use a more established and reviewed library for security critical applications. Open an issue or email me if you see any security issue or risk with this library.

The initial release of this package was targeted at python2.7. Earlier versions may work but have no guarantee of correctness or stability. As of release 1.2.1+ python3 is now supported as well.

Name Class Proposed By
P192 / secp192r1 fastecdsa.curve.P192 NIST / NSA
P224 / secp224r1 fastecdsa.curve.P224 NIST / NSA
P256 / secp256r1 fastecdsa.curve.P256 NIST / NSA
P384 / secp384r1 fastecdsa.curve.P384 NIST / NSA
P521 / secp521r1 fastecdsa.curve.P521 NIST / NSA
secp192k1 fastecdsa.curve.secp192k1 Certicom
secp224k1 fastecdsa.curve.secp224k1 Certicom
secp256k1 (bitcoin curve) fastecdsa.curve.secp256k1 Certicom
brainpoolP160r1 fastecdsa.curve.brainpoolP160r1 BSI
brainpoolP192r1 fastecdsa.curve.brainpoolP192r1 BSI
brainpoolP224r1 fastecdsa.curve.brainpoolP224r1 BSI
brainpoolP256r1 fastecdsa.curve.brainpoolP256r1 BSI
brainpoolP320r1 fastecdsa.curve.brainpoolP320r1 BSI
brainpoolP384r1 fastecdsa.curve.brainpoolP384r1 BSI
brainpoolP512r1 fastecdsa.curve.brainpoolP512r1 BSI

As of version 1.5.1 construction of arbitrary curves in Weierstrass form (y^2 = x^3 + ax + b (mod p)) is supported. I advise against using custom curves for any security critical applications. It's up to you to make sure that the parameters you pass here are correct, no validation of the base point is done, and in general no sanity checks are done. Use at your own risk.

from fastecdsa.curve import Curve
curve = Curve(
    name,  # (str): The name of the curve
    p,  # (long): The value of p in the curve equation.
    a,  # (long): The value of a in the curve equation.
    b,  # (long): The value of b in the curve equation.
    q,  # (long): The order of the base point of the curve.
    gx,  # (long): The x coordinate of the base point of the curve.
    gy,  # (long): The y coordinate of the base point of the curve.
    oid  # (str): The object identifier of the curve (optional).
)

Any hash function in the hashlib module (md5, sha1, sha224, sha256, sha384, sha512) will work, as will any hash function that implements the same interface / core functionality as the those in hashlib. For instance, if you wish to use SHA3 as the hash function the pysha3 package will work with this library as long as it is at version >=1.0b1 (as previous versions didn't work with the hmac module which is used in nonce generation).

Currently it does basic point multiplication significantly faster than the ecdsa package. You can see the times for 1,000 signature and verification operations below, fast.py corresponding to this package and regular.py corresponding to ecdsa package.

http://i.imgur.com/oNOfnG6.png?1

As you can see, this package in this case is ~25x faster.

If you'd like to benchmark performance on your machine you can do so using the command:

$ python setup.py benchmark

This will use the timeit module to benchmark 1000 signature and verification operations for each curve supported by this package.

You can use pip: $ pip install fastecdsa or clone the repo and use $ python setup.py install. Note that you need to have a C compiler. You also need to have GMP on your system as the underlying C code in this package includes the gmp.h header (and links against gmp via the -lgmp flag). On debian you can install all dependencies as follows:

$ sudo apt-get install python-dev libgmp3-dev

You can use this package to generate keys if you like. Recall that private keys on elliptic curves are integers, and public keys are points i.e. integer pairs.

from fastecdsa import keys, curve

# generate a private key for curve P256
priv_key = keys.gen_private_key(curve.P256)

# get the public key corresponding to the private key we just generated
pub_key = keys.get_public_key(priv_key, curve.P256)

Some basic usage is shown below:

from fastecdsa import curve, ecdsa, keys
from hashlib import sha384

m = "a message to sign via ECDSA"  # some message

''' use default curve and hash function (P256 and SHA2) '''
private_key = keys.gen_private_key(curve.P256)
public_key = keys.get_public_key(private_key, curve.P256)
# standard signature, returns two integers
r, s = ecdsa.sign(m, private_key)
# should return True as the signature we just generated is valid.
valid = ecdsa.verify((r, s), m, public_key)

''' specify a different hash function to use with ECDSA '''
r, s = ecdsa.sign(m, private_key, hashfunc=sha384)
valid = ecdsa.verify((r, s), m, public_key, hashfunc=sha384)

''' specify a different curve to use with ECDSA '''
private_key = keys.gen_private_key(curve.P224)
public_key = keys.get_public_key(private_key, curve.P224)
r, s = ecdsa.sign(m, private_key, curve=curve.P224)
valid = ecdsa.verify((r, s), m, public_key, curve=curve.P224)

''' using SHA3 via pysha3>=1.0b1 package '''
import sha3  # pip install [--user] pysha3==1.0b1
from hashlib import sha3_256
private_key, public_key = keys.gen_keypair(curve.P256)
r, s = ecdsa.sign(m, private_key, hashfunc=sha3_256)
valid = ecdsa.verify((r, s), m, public_key, hashfunc=sha3_256)

The Point class allows arbitrary arithmetic to be performed over curves. The two main operations are point addition and point multiplication (by a scalar) which can be done via the standard python operators (+ and * respectively):

# example taken from the document below (section 4.3.2):
# https://koclab.cs.ucsb.edu/teaching/cren/docs/w02/nist-routines.pdf

from fastecdsa.curve import P256
from fastecdsa.point import Point

xs = 0xde2444bebc8d36e682edd27e0f271508617519b3221a8fa0b77cab3989da97c9
ys = 0xc093ae7ff36e5380fc01a5aad1e66659702de80f53cec576b6350b243042a256
S = Point(xs, ys, curve=P256)

xt = 0x55a8b00f8da1d44e62f6b3b25316212e39540dc861c89575bb8cf92e35e0986b
yt = 0x5421c3209c2d6c704835d82ac4c3dd90f61a8a52598b9e7ab656e9d8c8b24316
T = Point(xt, yt, curve=P256)

# Point Addition
R = S + T

# Point Subtraction: (xs, ys) - (xt, yt) = (xs, ys) + (xt, -yt)
R = S - T

# Point Doubling
R = S + S  # produces the same value as the operation below
R = 2 * S  # S * 2 works fine too i.e. order doesn't matter

d = 0xc51e4753afdec1e6b6c6a5b992f43f8dd0c7a8933072708b6522468b2ffb06fd

# Scalar Multiplication
R = d * S  # S * d works fine too i.e. order doesn't matter

e = 0xd37f628ece72a462f0145cbefe3f0b355ee8332d37acdd83a358016aea029db7

# Joint Scalar Multiplication
R = d * S + e * T

Thanks to those below for contributing improvements:

  • targon