🚨🚨Yesterday one signature moved ~$6.3B out of
#Binance in about 20 minutes.
Biggest crypto hack in history. ~4× the
#Bybit heist
Here’s why the same breach of a
@kaspaunchained L1 covenant vault would have left the attacker with almost nothing. 🧵
2 - What actually happened (Binance + on-chain analysis):
→ Compromised DevOps engineer’s laptop
→ Attacker reached a service container that could make Binance’s hot-wallet signing system sign
→ No private key “leaked.” The signer just signed what it was told.
3 - On Ethereum the attacker didn’t even need to drain manually.
They sprayed 50 EIP-7702 authorisations at hot wallet “Binance 14” in one tx (nonces 18,914,775–824). One matched the live nonce.
That single signature handed the wallet to an attacker-deployed sweeper. ~3 minutes later: 167k ETH + 1.6B USDT + dozens of other tokens gone.
4 - Then the same signing system signed plain transfers on BTC, TRON, XRP, SOL, Arbitrum, Base, Avalanche, Optimism, Polygon.
10 chains. Different key types. Same blast radius.
Fee to move billions: under a dollar.
5 - The root problem isn’t Ethereum or Binance’s engineers.
It’s the model: in a hot wallet, the key is total authority.
Every “withdrawal limit,” every “approval tier,” every “wait 24h” lives on a server. If you own the signer, you own the policy.
6 - Now the Kaspa side.
Since the Toccata hard fork (30 June 2026), Kaspa L1 has covenants: a UTXO that inspects the transaction spending it (amounts, outputs, locktime, DAA score, its own successor).
The rules live inside the coins, not on a server. Enforced by every node.
7 - A covenant vault (written in Silverscript) can look like this:
→ Max X KAS per withdrawal
→ Only to a pre-fixed payout address
→ Only after N blocks / DAA score since last withdrawal
→ Recovery/alarm key can cancel or sweep to cold at any time
→ Output must be the same vault with updated state
8 - Replay the Binance breach against that vault.
Attacker steals the hot key. Signs a withdrawal. Signature is valid.
So what. The covenant checks the tx:
• Over the cap → invalid
• Wrong address → invalid
• Too soon → invalid
They get one capped tranche. Then the chain itself says “wait.”
9 - During that wait, your recovery key (cold, different machine, different person) sweeps everything to safety.
The attacker can’t stop it. The attacker can’t speed up the clock.
Because the clock isn’t a clock. It’s DAA score — Kaspa’s PoW block counter at ~10 blocks/sec.
10 - To skip the delay you’d have to mine blocks faster than the entire network.
Not phish them. Not bribe them. Out-hash the world.
That’s the only way to move the time forward. A pure PoW BlockDAG makes rewriting history computationally infeasible.
11 - And the 7702 trick? Doesn’t exist here.
There’s no opcode that lets a key hand an address over to arbitrary code. A UTXO’s spend rule is fixed the moment it’s created. You can spend it within the rules or not at all.
A stolen key can’t rewrite what the coins are allowed to do.
12 - Nothing described is theoretical.
KIP-10 and KIP-17 introspection are live on mainnet. Silverscript 1.0 shipped (Sept 2026). Covenant vaults with capped withdrawals + recovery paths are already deployed and verifiable by covenant ID.
This is the first L1 where “smart vault” means consensus-enforced, PoW-secured, UTXO-native.
13 - Binance will cover the losses. Stablecoin issuers froze billions. North Korea 🇰🇵 gets the rest.
Next time, ask the exchange holding your coins:
Is your withdrawal policy a server config, or does every node on a proof-of-work chain enforce it?
Only one of those survives a compromised laptop.
#Kaspa #KAS #Covenants #NotYourKeys