Syscoin Bridge
Why it happened
Attack class: Bridge exploitA malformed burn was read differently by Syscoin Core and the bridge relay, which released 5 billion SYS that were never burned.
What would have stopped it
- Make the relay reject any burn that is not in one canonical form.
- Test both parsers against each other on malformed inputs.
Not scored before the exploit
BlackHart had not scored Syscoin Bridge before this exploit.
An attacker tricked the Syscoin bridge into releasing 5 billion SYS, valued at about $10 million on paper, without burning any real coins first. A malformed transaction on Syscoin's UTXO chain was read one way by Syscoin Core and another way by the bridge relay, which treated it as a burn of native SYS. The amount was far larger than Syscoin's circulating supply, so it could not have been sold anywhere near that value, and no user balances were taken. Syscoin paused the bridge and traced the coins, the attacker returned all 5 billion SYS three days later, and Syscoin burned them to restore the normal supply.
A malformed burn on the UTXO chain was accepted by the NEVM relay, and the bridge vault released 5 billion SYS to the attacker's NEVM address, which moved them back to the UTXO chain before returning them all to the recovery address.
Attack flow
All 5 on-chain stepsThe attack in 3 steps. Step 1, Misread burn: Bridge vault on NEVM (bridge). Step 2, Bridge back: Attacker NEVM address (attacker), about $10M reached this step. Step 3, Returned: Syscoin recovery address (exit).
Step 1: Misread burn
- BridgeBridge vault on NEVM5,000,000,000 SYS released~ $10M notional0x7904299b...aCBd5f
Step 2: Bridge back
- AttackerAttacker NEVM address5B SYS sent back to the UTXO chainsplit into 4B and 1B0x71b216BD...554769
Step 3: Returned
- ExitSyscoin recovery addressall 5B SYS returned June 10burned June 15
Why each step worked
- Core and relay parsed one burn two waysrelay treated a worthless token burn as native SYS
- No cap on a single releasemissing controlone relay moved many times the circulating supply
- Bridge paused and coins tracedexchanges and partners blocked the funds
- Relay now rejects ambiguous burnsmatching Core change in progress
Full forensic detail
Step-by-step reconstruction, root cause, counterfactuals, remediation, and disclosure timeline.
Exploit anatomy
Root cause
Syscoin's bridge moves SYS between its Bitcoin-style UTXO chain and its EVM layer, NEVM. When a burn happens on one side, a relay contract on NEVM verifies proof of that burn and tells the vault to release the matching amount. Per Syscoin's technical post-mortem, the root cause was a cross-layer interpretation mismatch: Syscoin Core and the NEVM relay parsed the same malformed burn transaction differently. Core treated it as a burn of the attacker's worthless token, while the relay read it as a burn of native SYS, and the vault, which treats that asset as native SYS, released 5 billion coins that had never been burned. Halborn and Syscoin both stress that the cryptography and the proof format were not broken; the failure was in parsing, where two parts of one bridge must agree exactly on what a transaction says. The release dwarfed Syscoin's circulating supply, so the roughly $10 million figure is a notional market value rather than an amount the attacker could have sold. The relay now rejects ambiguous burns, and Syscoin Core is being updated to reject the same pattern consistently.
Prevention analysis
The attack needed Core and the relay to read one transaction two ways. Rejecting ambiguous burns outright, which Syscoin has now deployed, removes the gap.
Differential testing of Core and the relay on deliberately odd transactions would have exposed that they disagreed long before an attacker did.
A single relay released 5 billion SYS, many times the circulating supply. A per-transaction and daily ceiling would have kept any mistake small.
Similar incidents
The Ethereum side of the Verus bridge read signed data differently from Verus nodes and paid out a transfer that never happened. Same root: two layers of one bridge disagreeing about the same data.
A flaw in how the bridge checked proofs let an attacker mint 2 million BNB from nothing. Same class: proof parsing, not cryptography, was the weak point.
Nomad released funds for messages that were never sent. Same outcome: the destination side paid out without a real burn or lock behind it.
Remediation
Timeline
Get your protocol scored across 12 dimensions, or request ongoing coverage.