Drift Protocol Hack (2026): How a Durable Nonce Play Nuked $285M

April 1, 2026. Solana DeFi is vibing. Drift is the perp DEX everyone points to when they say Solana is actually usable. Then a bot alarm hits: funds are leaving vaults fast. Like, very fast. And Drift has to post the most cursed sentence in crypto history: This is not an April Fool's joke.

The result: roughly $285M gone from Drift Protocol, in an exploit that wasn't a spicy smart-contract bug. It was a privileged access faceplant: durable nonces, blind signing, multisig ops, and an attacker who turned a fake token into real withdrawals.

This post is the full breakdown  what happened, what durable nonces even are, why just don't sign shady stuff is not enough, and the practical steps you can take today (even if you're not a protocol dev) to avoid being exit liquidity in the next governance-speedrun.

Degen Intel

If you only read one thing: durable nonces let a transaction be signed now and executed later. If you blind-sign a durable nonce transaction, you might be approving future-you getting rugged  and you can't undo it by waiting for it to expire.

What the Drift Protocol hack was (in one paragraph)

On April 1, 2026, an attacker gained administrative control of Drift Protocol and drained about $285M (over 50% of its TVL) by combining (1) durable nonce-based delayed execution, (2) social engineering / blind signing to get legitimate multisig approvals, and (3) oracle + collateral manipulation using a fake token called CarbonVote Token (CVT) that was made to look like a real $1 asset with tiny liquidity and wash trading.

The timeline: months of setup, seconds of takeover

This wasn't a run a script, find a bug kind of hit. It was a slow-burn op with a fast execute.

Phase 1: Build the fake collateral

  • March 12, 2026: The attacker created CVT with a total supply of 750 million tokens.
  • They seeded a tiny liquidity pool (think pocket change money) and wash-traded it so price oracles saw a clean-ish ~$1 history.

Phase 2: Durable nonce + multisig approval trap

  • March 2330, 2026: durable nonce accounts were created and transactions were prepared for delayed execution.
  • March 26, 2026: Drift migrated to a new 2-of-5 Security Council multisig configuration with zero timelock.
  • April 1, 2026 at ~16:05 UTC: two transactions executed one second apart transferred admin control to an attacker-controlled address, then the draining began.

Chainalysis notes the admin-transfer sequence started at 16:05:18 UTC and executed at 16:05:19 UTC, moving admin control to H7PiGqqUaanBovwKgEtreJbKmQe6dbq6VTrw6guy7ZgL.

Durable nonces explained (why this attack worked)

Normally on Solana, transactions depend on a recent blockhash. If you sign something and it doesn't land quickly, it expires. That expiration window is a low-key safety feature.

Durable nonces replace that short-lived blockhash with a nonce stored in an on-chain account. The signature can stay valid indefinitely until the nonce advances. That means an attacker can:

  1. Get you to sign something when you're chill and distracted.
  2. Wait for the perfect moment (team asleep, liquidity high, monitoring blind).
  3. Execute the signed transaction later  as if you just approved it.

BlockSec calls this out directly: durable nonces decouple signer intent from execution timing, and signatures don't just time out the way people assume in day-to-day multisig ops.

How the money got drained: CVT as collateral, real assets out

After getting admin control, the attacker reconfigured the protocol like it was their own sandbox:

  • Whitelisted CVT as usable collateral.
  • Loosened risk parameters and set borrowing / withdrawal constraints absurdly high.
  • Deposited a massive amount of fake collateral (Chainalysis cites 500 million CVT).
  • Withdrew real assets against it across vaults.

Per Chainalysis, at least 18 assets were hit. The biggest chunks included $159.3M in JLP and $71.4M in USDC, plus cbBTC, USDT, USDS, WETH, dSOL, WBTC, FARTCOIN, and JitoSOL.

Why oracle manipulation matters even when the contract is audited

Most people hear oracle attack and imagine some complex hack. In practice, it can be painfully basic: if an oracle uses on-chain price history, and you can fake that history cheaply, you can fake value.

In this incident, a token with tiny liquidity was made to look stable and priced near $1. Once admin privileges let the attacker list it and loosen limits, the protocol started treating Monopoly money like collateral.

Where the funds went: bridging fast, laundering faster

Once the withdrawals started, time-to-exit was the whole game. Chainalysis observed the first bridge arrival to Ethereum about 23 minutes after the admin takeover, around 16:28 UTC. The drain continued for ~2.5 hours, with the last confirmed drain around 18:31 UTC.

TRM Labs also notes that most stolen funds were bridged to Ethereum within hours, with large USDC transfers that moved hundreds of thousands to millions per transaction.

Who did it? What we can say without coping

Attribution in crypto is always messy because attackers want it that way. Still: multiple investigators flagged indicators consistent with DPRK-linked operations. Elliptic says it saw multiple indicators linking the exploit to DPRK methods, and Drift later stated with medium-high confidence that the operation aligned with actors tied to the 2024 Radiant Capital hack attributed to UNC4736.

Translation: even if you ignore the state actor narrative, the operational maturity here is real. This wasn't a teen with Metasploit. This was a patient op that understood governance, ops, and human psychology.

How to not get rekt: the user + builder checklist

You can't personally patch Drift. But you can stop being the easy mode target.

If you're a normal user (trader, LP, degen)

  • Assume smart contract audits dont cover ops risk. The Drift exploit wasnt a classic code bug; it was admin + signing risk.
  • Watch for privileged changes. If a protocol can list new collateral or change withdrawal limits instantly, thats a risk factor. No timelock = no time to react.
  • Dont park life savings in one protocol. Split risk. Its boring. It works.
  • Prefer platforms where you control exposure. If you want to trade majors without being glued to a single DeFi venues governance risk, consider using a regulated platform for spot trading and keeping DeFi exposure intentional. For a clean multi-asset dashboard, Traderise is one option to keep your trading stack simpler.

If youre a protocol team / multisig signer (the real lesson)

  • Ban blind signing. If signers cant decode what theyre approving, youre already cooked.
  • Treat durable nonce transactions like loaded weapons. If you allow them, you need explicit process controls and monitoring.
  • Timelocks arent optional. A timelock converts a catastrophic instant admin change into a we have 12 hours to respond event.
  • Collateral listings need circuit breakers. Minimum liquidity, capped borrowing, staged rollout, and independent oracle sanity checks.
  • Operational security is product security. The attack surface includes conferences, DMs, repos, and quick sign this Slack messages.

Want to trade without juggling five apps?

If youre building a crypto portfolio and want a cleaner home base for trading (crypto + more), check out Traderise. Keep DeFi exposure intentional, not accidental.

Explore Traderise

Bottom line: this wasnt a hack, it was an approval

The most brutal part of the Drift Protocol hack is that the network did what it was told. The admin transfer transactions had valid signatures. The system treated them as legitimate. And once admin control flipped, everything after that was just normal functionality used maliciously.

So yeah, learn the technical details. But also learn the meta-lesson: in DeFi, the real boss fight is permissions. If you dont know who can change what (and how fast), youre not doing risk management  youre gambling.


Sources: Chainalysis breakdown of the Drift hack (timeline, durable nonces, stolen-asset breakdown) at https://www.chainalysis.com/blog/lessons-from-the-drift-hack/  Elliptic analysis (amount stolen estimate, TVL impact, DPRK indicators) at https://www.elliptic.co/blog/drift-protocol-exploited-for-286-million-in-suspected-dprk-linked-attack  BlockSec incident analysis (durable nonce explanation and phases) at https://blocksec.com/blog/drift-protocol-incident-multisig-governance-compromise-via-durable-nonce-exploitation  TRM Labs analysis (staging start and bridging pattern) at https://www.trmlabs.com/resources/blog/north-korean-hackers-attack-drift-protocol-in-285-million-heist