Lighter Exploit: Reused Nonces Exposed Trader API Keys

Quick answer. On November 16, 2026, about $10.4 million in USDC left Lighter after attackers recovered a trader's API keys. Lighter blames custom signing code that reused nonces, not its protocol. Its Security Council returned about $5 million; the final loss is unconfirmed.
Video transcript
On November 16, attackers recovered a Lighter trader's API keys and withdrew millions. Lighter blames custom signing code that reused nonces, not its protocol.
Bitquery saw 10.4 million dollars in USDC leave Lighter in one Ethereum withdrawal, an observed outflow, not a confirmed loss.
A signature mixes the private key with a one-time secret, the nonce. Reuse a nonce across two messages, and the key can be solved. Lighter's API signatures were public.
The attacker created a public pool from the affected account, bought shares from another account, moved the funds in, and redeemed them. Each step used normal account features.
Lighter's onchain safeguard paused withdrawals and new pool creation. Its Security Council returned about 5 million dollars to the trader, and withdrawals resumed.
Lighter says two changes are live: non-public API signatures, and wallet signatures for pool transfers. IP allowlists, per-key limits and nonce-reuse monitoring are planned, with no dates.
Still open: the final loss, how many accounts were hit, and when pool creation reopens. A 15.5 million dollar estimate across two accounts is unconfirmed.
For API key users: use the official signer, test that nonces never repeat, map every route a key can sign for, and set per-key limits.
A contract audit would not read a trader's signing code. Read the full article, or talk to Guardian about pentesting signing, key scoping and revocation.
What Happened
On November 16, 2026, at 10:11 UTC, 10,393,057 USDC left Lighter's Ethereum settlement contract in a single withdrawal, Bitquery observed. Bitquery
Lighter published a detailed postmortem at about 18:20 UTC the same day. Lighter, Cointelegraph
It says attackers recovered a trader's API private keys from public signatures. The trader's custom or third-party signing code had reused cryptographic nonces. Coinfolytics
Lighter says its protocol, smart contracts, infrastructure and official SDKs were not affected.
How Reused Nonces Leak Keys
Schnorr and ECDSA signatures, including the Schnorr scheme Lighter's signers use, combine the private key with a secret one-time number, the nonce. Certicom Research / Standards for Efficient Cryptography Group, IETF, Bitcoin Improvement Proposals, Elliot Technologies
If two signatures over different messages share a nonce, anyone holding both can solve for the private key with simple algebra. National Institute of Standards and Technology
That signing nonce is not the transaction nonce Lighter's official SDK increments per API key to order transactions.
Visibility made that practical here. Lighter's API-key signatures were public, so repeated nonces turned published data into recoverable keys. Lighter, Cointelegraph, Coinfolytics
RFC 6979 warns that even slight bias in how nonces are generated can be turned into attacks. Each signature needs a fresh, unpredictable nonce.
From API Key to Withdrawal
Lighter's architecture documentation says exchange operations are user-signed, and that public pools are part of its Ethereum-anchored state. Lighter
Its public-pool documentation says whitelisted operators trade pooled capital through sub-accounts. Participants receive shares they redeem when they withdraw.
The reported route used those ordinary functions. The attacker created a public pool from the affected account, then bought pool shares from a separate account. Coinfolytics
The attacker moved the affected funds into the pool and redeemed the shares, taking the value out.
The failure was an offchain signer, followed by abuse of normal account features.
Guardian analysis: a key's real power is the sum of every route it can sign for. A trading key that can fund a pool can, in effect, move money.
The 3Commas API key leak showed a related pattern: stolen trading keys used to drain balances through counter-trades rather than transfers.
The Onchain Trail
Within minutes of the 10:11 UTC withdrawal, about 10.36 million USDC was swapped into ETH, Bitquery traced. Bitquery, Cointelegraph
Between 10:19 and 11:55 UTC, twelve new wallets received 160 ETH each, 1,920 ETH in total.
Each sent its ETH through LI.FI within three minutes. About 16 to 18 minutes later, a matching Hyperliquid account bought XMR1.
In all, 6,705 XMR1 was bought for about $3.46 million. At 12:44 UTC, 6,700 XMR1 was still held.
At Bitquery's 11:55 UTC snapshot, 3,845 ETH was still in the original wallet, worth about $7 million at that time.
That is a point-in-time observation, not a loss or recovery figure. Bitquery says it did not check what backs XMR1.
Bitquery's trace stops at Ethereum. It could not inspect Lighter's rollup, so the pool steps rest on Lighter's own account.
Pause and Recovery
Lighter's onchain safeguard, which it calls Guardian and which has no connection to Guardian, the publisher of this hub, detected the anomaly. It paused withdrawals and new pool creation. Lighter, Cointelegraph, Coinfolytics
Lighter's Security Council then returned about $5 million in USDC to the affected trader. Coin Edition
Lighter then switched off its Guardian mode, and withdrawals resumed.
Guardian analysis: the pause bought time, but the recovery relied on a council able to change state. Traders on any venue should know who holds that power and how it is checked.
Our guide to emergency pause design covers who should hold pause and unpause powers, and which paths should stay open.
What Lighter Changed
Lighter says two changes are live. It has given no dates for the rest. Lighter, Cointelegraph, Coinfolytics, Coin Edition, Daily Crypto Briefs, TokenPost
It told anyone who used custom or third-party signing code to rotate their API keys and move to an official SDK.
Guardian analysis: the wallet-signature rule matters most. On Lighter's description, a stolen API key alone should no longer be able to fund a public pool.
| Control | Status per Lighter |
|---|---|
| API-key signatures made non-public | Done |
| Wallet signature required for transfers into public pools | Done |
| Voiding affected keys and contacting affected accounts | Planned, no date given |
| IP allowlists for API access | Planned, no date given |
| Configurable per-key permissions and limits | Planned, no date given |
| Automated nonce-reuse monitoring | Planned, no date given |
What Is Still Unknown
Lighter's postmortem is posted on X. Its contents here are taken from independent reports of it. Lighter
- Final loss: Bitquery saw $10.4 million withdrawn and cites X posts estimating $15.5 million across two market-maker accounts, unconfirmed by Lighter. Bitquery, Cointelegraph, Coinfolytics, Coin Edition, Daily Crypto Briefs
- How many accounts and keys were affected. Lighter refers to one trader; other reports refer to two market-maker accounts.
- A full reconciliation of the roughly $5 million returned against the $10.39 million Ethereum withdrawal.
- The recovery itself. Lighter has published no Security Council transaction or upgrade code for it. DeFiPrime
- When new public-pool creation reopens. Withdrawals resumed on November 16, but Lighter gave no timetable for pools.
Checklist for Teams Using API Keys
These checks apply to any venue where an API key signs orders, transfers or pool actions.
Our key rotation best practices guide covers overlapping and revoking API keys without downtime.
- Do not write your own nonce generation; use the signing path the venue maintains.
- Test any signer you maintain: sign many messages and assert that no nonce ever repeats.
- Treat every signature you publish as permanent and public.
- Map every route a key can sign for, including pools, sub-accounts and internal transfers.
- Give each key only the permissions and limits its strategy needs, and bind it to known IP addresses.
- Alert on new pools, unusual transfers and withdrawals from trading accounts.
- Rehearse revoking all keys at once, and confirm the venue's emergency contact path.
Guardian Analysis: What the Portfolio Shows
A case-insensitive search of 321 current public Guardian reports for "API key" matched 3 reports, containing 100 findings in total. The count covers reports that mention API keys, not a rate of exposure or loss. Guardian
The closest match is Synthetix L-35. An Alchemy API key exposed in test cases could let anyone make calls from the account; the finding is marked resolved.
The fix review (L-05, Low, acknowledged) found the key, moved to an environment variable, still visible in commit history. Guardian recommended deprecating it, since moving a key does not retire it.
L-35 is credential exposure, not nonce reuse, and not evidence of fund movement. Lighter shows why key exposure, signer behavior and key permissions belong in one test.
GMX L-03 (Low, acknowledged) set out safeguards against draining GMX's Gelato relay balance: a capped balance, requests routed through a backend holding the private Gelato API key, rate limits, balance monitoring with auto-pause, and whitelisting.
Those are the same kinds of limits Lighter now plans per key. They cap what a stolen credential is worth.
Searches for reused nonce and nonce reuse matched none of the 321 current public reports, so the portfolio supports no prevalence figure for this failure.
A contract audit would not read a trader's signing code. Our offchain automation pentest guide covers testing bot keys, signing services and what each key can call.
Frequently Asked Questions
How quickly can Lighter public pool shares be redeemed?
Lighter's documentation says public pools have no lockup periods, only a 5-minute cooldown on Lighter. So a share bought from one account can be redeemed within minutes. That puts the weight on detection and on who can authorize transfers into a pool. Lighter
What should I do if I am unsure which signer my bot uses?
Lighter urged users unsure whether their signing code reused a nonce to rotate all their API keys. Then review account history for pools, pool shares or transfers you did not create. Coin Edition
Related reading
- 3commas api key leak explained
- crypto exchange pentesting
- offchain automation pentest
- crypto key rotation
- web3 sdk security audit
Sources
- Lighter: Signing Safely on Lighter: Lessons from November 16
- Lighter: Public Pools
- Lighter: Technical Architecture: Lighter Core
- Bitquery: Lighter Incident: $10.4M Withdrawn, Leaving 160 ETH at a Time
- Cointelegraph: Lighter recovers $5M after $10M exploit linked to compromised API keys
- Coinfolytics: Lighter resumes withdrawals after recovering $5 million following API key exploit
- Coin Edition: Lighter Restores Withdrawals After $5M USDC Exploit Recovery
- Guardian: Alchemy API Key Exposed, Guardian finding L-35
- Guardian: Rootstock Deployment
- Guardian: Gasless Sponsored Calls, Part 1
- Guardian: Blockchain Penetration Testing Services
- Daily Crypto Briefs: Lighter Resumes Withdrawals, Returns $5 Million in USDC
- DeFiPrime: Lighter pauses withdrawals after trader API key leak
- TokenPost: Lighter Returns About $5 Million in USDC and Restarts Withdrawals
- IETF: RFC 6979: Deterministic Usage of the Digital Signature Algorithm (DSA) and Elliptic Curve Digital Signature Algorithm (ECDSA)
- Certicom Research / Standards for Efficient Cryptography Group: SEC 1: Elliptic Curve Cryptography, Version 2.0
- Elliot Technologies: lighter-go: Public Go SDK for Lighter
- Elliot Technologies: poseidon_crypto Schnorr signer implementation
- Elliot Technologies: lighter-go module dependencies
- Bitcoin Improvement Proposals: BIP 340: Schnorr Signatures for secp256k1
- National Institute of Standards and Technology: Notes on Threshold EdDSA/Schnorr Signatures
Test the signer behind your trading keys
Guardian's blockchain penetration testing covers wallet signing, API key scoping and revocation, value-moving flows and key management: the offchain surface where this incident sat.