Mercado Bitcoin hack: how one signing breach drained $46 million across ten blockchains

The CyberSec Guru

Mercado Bitcoin hack

If you like this post, then please share it:

Buy me A Coffee!

Support The CyberSec Guru’s Mission

🔐 Fuel the cybersecurity crusade by buying me a coffee! Why your support matters: Zero paywalls: Keep the main content 100% free for learners worldwide.

“Your coffee keeps the servers running and the knowledge flowing in our fight against cybercrime.”☕ Support My Work

Buy Me a Coffee Button

The Mercado Bitcoin hack began overnight into Monday, November 2, 2026. By morning, Brazil’s largest cryptocurrency exchange had suspended all deposits and withdrawals after unauthorized outbound transfers hit its wallets. Arkham Intelligence later tracked the attacker’s holdings to about $46 million across 36 addresses on at least ten blockchains, and every transfer carried a valid signature.

The exchange’s first statement, posted in Portuguese on X, said engineers had identified and quickly contained an incident in an internal system. It added that no customer balances were affected and no personal data was exposed. The on-chain data showed a larger problem. Hypernative flagged more than $40 million in suspicious outflows from wallets attributed to Mercado Bitcoin. The attacker did not exploit a smart contract or a cross-chain bridge. They made ordinary, cryptographically valid transfers on several networks at the same time. I read that as a failure at the signing layer: once the middleware that signs transactions or orchestrates keys is breached, the cryptographic separation between independent blockchains stops protecting anything.

What happened at Mercado Bitcoin on November 2

On-chain forensics put the start of the drain at around 02:15 UTC, with sweeps from legacy Bitcoin deposit addresses tied to the exchange. Data compiled by Hypernative and Arkham analyst Emmett Gallic shows the attacker emptied accessible operational and deposit wallets on Bitcoin, Ethereum, Solana, Tron, Stellar, Avalanche, Polygon, Arbitrum, Optimism, Base, BNB Smart Chain and the XDC Network. That list has twelve names while reports speak of at least ten, so treat the exact network count as unsettled.

In the hours after the first extraction, Gallic noted that the attacker kept sweeping older, dormant deposit addresses. Smaller altcoins and meme tokens were dumped into decentralized liquidity pools to consolidate the value into native chain assets.

Mercado Bitcoin’s statement versus the on-chain data

When the exchange first halted transfers, it gave no dollar figures, no list of affected networks and no date for resuming withdrawals. It said it would cover any operational shortfall from corporate reserves, and Brazilian outlet Exame put the local-currency equivalent of the initial estimate at roughly R$ 217.7 million. Only after independent researchers began circulating transaction hashes did Mercado Bitcoin publish five primary destination addresses, shared “exclusively for identification and tracking purposes,” covering EVM networks, Bitcoin, Solana, Stellar and Tron.

The gap between Hypernative’s $40 million alert and Arkham’s $46 million valuation comes from how trackers count. One was following sweeps still in progress, the other a static snapshot. The attacker held signing credentials for historical deposit addresses, so automated scripts kept pulling residual balances even after the main hot wallets were empty.

The company’s claim that customer balances are untouched deserves a closer look. Mercado Bitcoin says its cold storage reserves back user liabilities one-to-one. Yet the on-chain evidence shows the compromised addresses included active customer deposit endpoints and operational liquidity pools. Customer funds are fungible in custodial accounting, so “no customer lost money” appears to mean the parent company chose to cover the deficit from its own balance sheet instead of cutting user accounts.

Where the stolen funds went

By midday on November 2, the destination clusters showed what had been taken from the treasury and operational reserves.

NetworkDestination addressAmount observed
Bitcoin19qGpsHvzh1t6pVi526XNNkh6hW6prDZp106.07 BTC (about $8.24 million), unspent at the first forensic check
Ethereum and EVM networks0xe485e7A70b1131216Cb47CA4555c4Bbb2C7424782,852.07 ETH on mainnet as the initial transfer
Solana3YX6WHncP1Ra4AotE9YLawFt4nPmxyux288FWXkEW97JAbout 10,700 SOL (around $1 million)
StellarGDXYRI6IONQA5YTGRLQ603FGJ2BWXPJTTQPUXFPSEXSRKBZLFKMZSOLPAbout 6.7 million XLM
TronTEX3HuDRHRa3tnYpy96M4UfSx62VhDyA5L1.7 million TRX plus substantial stablecoin balances

Across Ethereum, Base, Optimism and Arbitrum, Arkham’s 36-address cluster held roughly 3,669 ETH ($8.37 million), $1.76 million in PEPE and hundreds of smaller ERC-20 positions. Hypernative’s alerts also captured 77,000 AVAX on Avalanche and 3.07 million XDC on the XDC Network.

How one breach signed transactions on ten blockchains

Every transaction that left the exchange’s wallets was a standard native transfer (MsgSend, eth_sendRawTransaction or a regular UTXO spend) signed with a valid key. No contract was manipulated. That matters because these networks do not share a signature scheme or a transaction format.

📬 Stay Ahead of Cyber Threats

Get the latest cybersecurity news, critical vulnerabilities, threat intelligence, tutorials, and exclusive giveaways delivered straight to your inbox. No spam. Unsubscribe anytime.

Subscribe to the Newsletter →

Bitcoin, Ethereum, BNB Smart Chain, Avalanche, Polygon and Tron use the secp256k1 curve, with ECDSA (or Schnorr for Bitcoin Taproot). Even among those chains, key derivation paths (BIP-32 and BIP-44), address encoding and payload structure differ. Bitcoin’s UTXO model needs script-sig or witness data for each input. Ethereum and other EVM chains use account-based state transitions with RLP-encoded payloads, a chain-specific chainId under EIP-155 and sequential nonces. Solana and Stellar use Ed25519, which is mathematically incompatible with secp256k1 keys unless both come from one master seed (BIP-39) in a unified derivation tree.

To get valid signatures on both curves, from active hot wallets and from older user deposit addresses, the compromise most plausibly sat at one of two layers.

Master seed or key management service exfiltration

Suppose the exchange derives its multi-chain hot wallets and deterministic deposit addresses (HD wallets) from a small set of master seeds held in cloud KMS instances such as AWS KMS or HashiCorp Vault, without strict hardware isolation. An attacker who breaches the cloud control plane or steals decryption credentials could rebuild the private keys offline. From there they can sign and broadcast sweeps at their own pace through local nodes or public RPC endpoints. That fits the old deposit addresses still being swept hours after Mercado Bitcoin said the incident was contained.

Signing orchestration and policy engine compromise

Most exchanges keep keys in hardware security modules (HSMs) or multi-party computation (MPC) clusters, such as Fireblocks, Copper or proprietary threshold schemes, so a raw private key never sits in one memory space. An internal withdrawal service authenticates with API keys or mTLS certificates and submits unsigned transaction payloads to the signer. If the policy engine does not enforce destination whitelisting, rate limits or multi-human quorum approval for operational sweeps, an attacker who controls the withdrawal orchestration server can feed any destination (0xe485e…, 19qGps…, 3YX6WH…) into the pipeline. The HSM or MPC nodes see a request from an authenticated internal service and produce valid secp256k1 and Ed25519 signatures on every connected chain.

Why the sweep looks automated

The attacker took everything: blue-chip BTC and ETH reserves, illiquid meme coins, and dust in forgotten deposit addresses. That points to programmatic access to the exchange’s wallet management database and signing API, or to direct export of the HD derivation seeds. The pattern fits an automated sweeper script that iterated through the exchange’s indexed address table, queried balances over ten RPC networks, and signed max-balance transfers to the five pre-generated consolidation addresses, rather than hand-built transactions.

Can stolen crypto be frozen? Tether’s USDT freeze and its limits

Once a valid transaction is confirmed in a block, Layer-1 protocols have no way to reverse it. The bitcoin in 19qGpsHvzh1t6pVi526XNNkh6hW6prDZp and the ether in 0xe485e7A70b1131216Cb47CA4555c4Bbb2C742478 belong, in practice, to whoever holds the matching private keys. The only immediate countermeasure during a live multi-chain drain sits at the smart contract layer, in token contracts that an issuer administers and that include blacklist functions.

Within hours of Mercado Bitcoin publishing the attacker’s addresses, Tether CEO Paolo Ardoino confirmed on X that Tether had frozen more than 4.5 million USDT linked to the breach. On EVM networks and Tron, the USDT contract has a privileged method, addBlackList(address _evilUser), controlled by Tether’s multi-signature treasury keys. Calling it sets isBlackListed[_evilUser] = true in the contract’s storage. The standard transfer and transferFrom functions check require(!isBlackListed[msg.sender]) before changing any balance, so the USDT sitting in 0xe485e7A70b1131216Cb47CA4555c4Bbb2C742478 and TEX3HuDRHRa3tnYpy96M4UfSx62VhDyA5L is stuck. The attacker still holds the private keys and can move native ETH or TRX, but any call to the USDT contract reverts.

Brazilian outlets first reported an early tranche of $3.7 million before the freeze total passed $4.5 million. Even so, nearly 90% of the $46 million Arkham tracked sat outside the reach of any contract-level freeze. BTC, ETH, SOL, XLM and AVAX have no central issuer and no blacklist mapping. Experienced attackers also expect stablecoin freezes. They often swap stolen ERC-20 tokens into native ETH through aggregators like 1inch or Uniswap within minutes of the drain, well before the exchange’s security team has verified the breach, published addresses and reached Tether’s compliance desk.

Mercado Bitcoin and the $6.36 billion Binance breach

The Mercado Bitcoin hack came four days after the largest cryptocurrency theft in history. On October 29, 2026, attackers the FBI and TRM Labs linked to North Korea’s TraderTraitor group drained about $6.36 billion from Binance across ten blockchains in roughly 20 minutes. The haul dwarfed the $1.5 billion Bybit breach of February 2025.

The Binance attack followed a similar multi-chain pattern. After compromising an engineer’s workstation to reach internal withdrawal signing systems, the attackers made ordinary signed transfers across XRP, Bitcoin, Solana, Optimism, Arbitrum and Avalanche. On Ethereum they used pre-signed EIP-7702 delegation authorizations to turn a Binance hot wallet into an automated sweeper contract. Afterward, Tether, Circle, Ripple, World Liberty Financial and the Arbitrum Security Council coordinated to freeze roughly $2.7 billion in stolen assets. The attackers still kept more than 420,000 ETH and 18,270 BTC, about $2.4 billion in assets no one can freeze.

Neither Mercado Bitcoin nor law enforcement has linked the Brazilian breach to the Binance attack or to TraderTraitor. What the two incidents share is a weakness in how centralized exchanges build multi-chain withdrawal pipelines.

How exchanges can harden multi-chain signing

Over the past three years, exchanges have added dozens of Layer-1 and Layer-2 networks to capture trading volume. To avoid the latency and admin overhead of an air-gapped signing ceremony for every network, engineering teams folded wallet orchestration into unified middleware that talks to every supported chain. If that layer treats internal API requests as trusted by default, or if deterministic deposit keys live in a shared software environment, the exchange’s whole security perimeter reduces to one choke point.

Fixing that means moving validation out of the application server and into the cryptographic hardware. HSM and MPC enclaves need their own hardware-level velocity limits, strict destination whitelisting for cold-to-hot sweeps, and out-of-band cryptographic quorum approval whenever outbound volume in any rolling window rises above normal withdrawal variance. Until exchanges remove the middleware’s ability to sign arbitrary transfers on ten networks at once, the same drain stays available to anyone who breaches it.

Buy me A Coffee!

Support The CyberSec Guru’s Mission

🔐 Fuel the cybersecurity crusade by buying me a coffee! Your contribution powers free tutorials, hands-on labs, and security resources.

Why your support matters:
  • Writeup Access: Get complete writeup access within 12 hours
  • Zero paywalls: Keep the main content 100% free for learners worldwide

Perks for one-time supporters:
☕️ $5: Shoutout in Buy Me a Coffee
🛡️ $8: Fast-track Access to Live Webinars
💻 $10: Vote on future tutorial topics + exclusive AMA access

“Your coffee keeps the servers running and the knowledge flowing in our fight against cybercrime.”☕ Support My Work

Buy Me a Coffee Button

If you like this post, then please share it:

News

Discover more from The CyberSec Guru

Subscribe to get the latest posts sent to your email!

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.