Struct ed25519_dalek::Keypair

pub struct Keypair {
    pub secret: SecretKey,
    pub public: PublicKey,
}

An ed25519 keypair.

Fields

secret: SecretKey

The secret half of this keypair.

public: PublicKey

The public half of this keypair.

Implementations

impl Keypair

pub fn to_bytes(&self) -> [u8; 64]

Convert this keypair to bytes.

Returns

An array of bytes, [u8; KEYPAIR_LENGTH]. The first SECRET_KEY_LENGTH of bytes is the SecretKey, and the next PUBLIC_KEY_LENGTH bytes is the PublicKey (the same as other libraries, such as Adam Langley's ed25519 Golang implementation).

pub fn from_bytes<'a>(bytes: &'a [u8]) -> Result<Keypair, SignatureError>

Construct a Keypair from the bytes of a PublicKey and SecretKey.

Inputs

• bytes: an &[u8] representing the scalar for the secret key, and a compressed Edwards-Y coordinate of a point on curve25519, both as bytes. (As obtained from Keypair::to_bytes().)

Warning

Absolutely no validation is done on the key. If you give this function bytes which do not represent a valid point, or which do not represent corresponding parts of the key, then your Keypair will be broken and it will be your fault.

Returns

A Result whose okay value is an EdDSA Keypair or whose error value is an SignatureError describing the error that occurred.

pub fn generate<R>(csprng: &mut R) -> Keypair where
    R: CryptoRng + RngCore, 

Generate an ed25519 keypair.

Example

extern crate rand;
extern crate ed25519_dalek;


use rand::rngs::OsRng;
use ed25519_dalek::Keypair;
use ed25519_dalek::Signature;

let mut csprng = OsRng{};
let keypair: Keypair = Keypair::generate(&mut csprng);

Input

A CSPRNG with a fill_bytes() method, e.g. rand_os::OsRng.

The caller must also supply a hash function which implements the Digest and Default traits, and which returns 512 bits of output. The standard hash function used for most ed25519 libraries is SHA-512, which is available with use sha2::Sha512 as in the example above. Other suitable hash functions include Keccak-512 and Blake2b-512.

pub fn sign_prehashed<D>(
    &self,
    prehashed_message: D,
    context: Option<&[u8]>
) -> Result<Signature, SignatureError> where
    D: Digest<OutputSize = U64>, 

Sign a prehashed_message with this Keypair using the Ed25519ph algorithm defined in RFC8032 §5.1.

Inputs

• prehashed_message is an instantiated hash digest with 512-bits of output which has had the message to be signed previously fed into its state.
• context is an optional context string, up to 255 bytes inclusive, which may be used to provide additional domain separation. If not set, this will default to an empty string.

Returns

An Ed25519ph [Signature] on the prehashed_message.

Examples

extern crate ed25519_dalek;
extern crate rand;

use ed25519_dalek::Digest;
use ed25519_dalek::Keypair;
use ed25519_dalek::Sha512;
use ed25519_dalek::Signature;
use rand::rngs::OsRng;

let mut csprng = OsRng{};
let keypair: Keypair = Keypair::generate(&mut csprng);
let message: &[u8] = b"All I want is to pet all of the dogs.";

// Create a hash digest object which we'll feed the message into:
let mut prehashed: Sha512 = Sha512::new();

prehashed.update(message);

If you want, you can optionally pass a "context". It is generally a good idea to choose a context and try to make it unique to your project and this specific usage of signatures.

For example, without this, if you were to convert your OpenPGP key to a Bitcoin key (just as an example, and also Don't Ever Do That) and someone tricked you into signing an "email" which was actually a Bitcoin transaction moving all your magic internet money to their address, it'd be a valid transaction.

By adding a context, this trick becomes impossible, because the context is concatenated into the hash, which is then signed. So, going with the previous example, if your bitcoin wallet used a context of "BitcoinWalletAppTxnSigning" and OpenPGP used a context (this is likely the least of their safety problems) of "GPGsCryptoIsntConstantTimeLol", then the signatures produced by both could never match the other, even if they signed the exact same message with the same key.

Let's add a context for good measure (remember, you'll want to choose your own!):

let context: &[u8] = b"Ed25519DalekSignPrehashedDoctest";

let sig: Signature = keypair.sign_prehashed(prehashed, Some(context))?;

pub fn verify(
    &self,
    message: &[u8],
    signature: &Signature
) -> Result<(), SignatureError>

Verify a signature on a message with this keypair's public key.

pub fn verify_prehashed<D>(
    &self,
    prehashed_message: D,
    context: Option<&[u8]>,
    signature: &Signature
) -> Result<(), SignatureError> where
    D: Digest<OutputSize = U64>, 

Verify a signature on a prehashed_message using the Ed25519ph algorithm.

Inputs

• prehashed_message is an instantiated hash digest with 512-bits of output which has had the message to be signed previously fed into its state.
• context is an optional context string, up to 255 bytes inclusive, which may be used to provide additional domain separation. If not set, this will default to an empty string.
• signature is a purported Ed25519ph [Signature] on the prehashed_message.

Returns

Returns true if the signature was a valid signature created by this Keypair on the prehashed_message.

Examples

extern crate ed25519_dalek;
extern crate rand;

use ed25519_dalek::Digest;
use ed25519_dalek::Keypair;
use ed25519_dalek::Signature;
use ed25519_dalek::SignatureError;
use ed25519_dalek::Sha512;
use rand::rngs::OsRng;

let mut csprng = OsRng{};
let keypair: Keypair = Keypair::generate(&mut csprng);
let message: &[u8] = b"All I want is to pet all of the dogs.";

let mut prehashed: Sha512 = Sha512::new();
prehashed.update(message);

let context: &[u8] = b"Ed25519DalekSignPrehashedDoctest";

let sig: Signature = keypair.sign_prehashed(prehashed, Some(context))?;

// The sha2::Sha512 struct doesn't implement Copy, so we'll have to create a new one:
let mut prehashed_again: Sha512 = Sha512::default();
prehashed_again.update(message);

let verified = keypair.public.verify_prehashed(prehashed_again, Some(context), &sig);

assert!(verified.is_ok());

pub fn verify_strict(
    &self,
    message: &[u8],
    signature: &Signature
) -> Result<(), SignatureError>

Strictly verify a signature on a message with this keypair's public key.

On The (Multiple) Sources of Malleability in Ed25519 Signatures

This version of verification is technically non-RFC8032 compliant. The following explains why.

1. Scalar Malleability

The authors of the RFC explicitly stated that verification of an ed25519 signature must fail if the scalar s is not properly reduced mod \ell:

To verify a signature on a message M using public key A, with F being 0 for Ed25519ctx, 1 for Ed25519ph, and if Ed25519ctx or Ed25519ph is being used, C being the context, first split the signature into two 32-octet halves. Decode the first half as a point R, and the second half as an integer S, in the range 0 <= s < L. Decode the public key A as point A'. If any of the decodings fail (including S being out of range), the signature is invalid.)

All verify_*() functions within ed25519-dalek perform this check.

2. Point malleability

The authors of the RFC added in a malleability check to step #3 in §5.1.7, for small torsion components in the R value of the signature, which is not strictly required, as they state:

Check the group equation [8][S]B = [8]R + [8][k]A'. It's sufficient, but not required, to instead check [S]B = R + [k]A'.

History of Malleability Checks

As originally defined (cf. the "Malleability" section in the README of this repo), ed25519 signatures didn't consider any form of malleability to be an issue. Later the scalar malleability was considered important. Still later, particularly with interests in cryptocurrency design and in unique identities (e.g. for Signal users, Tor onion services, etc.), the group element malleability became a concern.

However, libraries had already been created to conform to the original definition. One well-used library in particular even implemented the group element malleability check, but only for batch verification! Which meant that even using the same library, a single signature could verify fine individually, but suddenly, when verifying it with a bunch of other signatures, the whole batch would fail!

"Strict" Verification

This method performs both of the above signature malleability checks.

It must be done as a separate method because one doesn't simply get to change the definition of a cryptographic primitive ten years after-the-fact with zero consideration for backwards compatibility in hardware and protocols which have it already have the older definition baked in.

Return

Returns Ok(()) if the signature is valid, and Err otherwise.

Trait Implementations

impl Debug for Keypair

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Signer<Signature> for Keypair

fn try_sign(&self, message: &[u8]) -> Result<Signature, SignatureError>

Sign a message with this keypair's secret key.

fn sign(&self, msg: &[u8]) -> S

Sign the given message and return a digital signature

impl Verifier<Signature> for Keypair

fn verify(
    &self,
    message: &[u8],
    signature: &Signature
) -> Result<(), SignatureError>

Verify a signature on a message with this keypair's public key.