Secret Network
Why it happened
Attack class: Bridge exploitThe Secret-side contract minted on any deposit naming a listed token, without checking which channel sent it.
What would have stopped it
- Authorize every mint by the packet's source channel
- Re-audit a fork when it changes the trust model
Not scored before the exploit
BlackHart had not scored Secret Network before this exploit.
An attacker drained about $4.67 million from the Axelar to Secret Network bridge by sending fake deposits from a chain they controlled to the Secret-side bridge contract, which minted seven kinds of Axelar-bridged tokens with nothing behind them. They then redeemed those tokens through the real bridge, and Axelar's escrow paid out its genuine USDT, USDC, ETH, WBTC, DAI, BNB and wstETH reserves within minutes. Only Axelar-bridged tokens on Secret lost their backing; SCRT, other Secret tokens, other IBC routes and Axelar itself were unaffected. The proceeds were swapped to ETH and cashed out mainly through ChangeNOW and KuCoin, and about $770,000 still sits in the attacker's Axelar wallet.
From a counterfeit chain, the attacker's Secret account relayed fake deposits into the Secret bridge contract, which minted unbacked tokens; redeeming them sent Axelar's real reserves to the attacker's Axelar wallet, and from there through Osmosis to an Ethereum wallet that cashed out on exchanges.
Attack flow
All 5 on-chain stepsThe attack in 4 steps. Step 1, Fake deposits: Secret bridge contract (bridge). Step 2, Escrow drain: Attacker Axelar wallet (attacker), about $4.67M reached this step. Step 3, Swap: Attacker Ethereum wallet (attacker), about $3.9M reached this step. Step 4, Cash out: ChangeNOW (exit), KuCoin (exit), about $1.8M reached this step.
Step 1: Fake deposits
- BridgeSecret bridge contract7 unbacked saToken mintsin about 6 minutessecret1yxj...sxee83
Step 2: Escrow drain
- AttackerAttacker Axelar walletUSDT, USDC, WETH, WBTC, DAI, WBNB, wstETH~ $4.67M releasedaxelar1hzr...hl6ad8
Step 3: Swap
- AttackerAttacker Ethereum walletvia Osmosis and Axelar GMP~ 2,349 ETH after swaps0x6c2eab82...123976
Step 4: Cash out
- ExitChangeNOW~ 1,050 ETHplus about 106 BNB0xeba88149...e94cb1
- ExitKuCoin~ 1,029 ETHvia deposit addresses0x45300136...92b785
Why each step worked
- Source channel never checkedallow list keyed by token name only
- Permissionless IBC channel openingcounterfeit chain connected on channel-227
- Fork shipped without re-auditmissing controlescrow checks removed when the contract was rebuilt to mint
- No outflow cap or anomaly pause on escrowmissing controlfull reserve paid to one wallet, unnoticed for 7 days
- Packet forwarding through Osmosisosmo1hzra9z4zn8q0w8f3dj2wnw0xgetu8dfdm2l9s5
- 30 fresh deposit wallets and spoof tokensused to clutter the trail
Full forensic detail
Step-by-step reconstruction, root cause, counterfactuals, remediation, and disclosure timeline.
Exploit anatomy
Root cause
The Secret-side contract for the Axelar connection (ics20-for-axelar, published by SCRT Labs and forked from its escrow-based snip20-ics20) minted Secret Axelar tokens whenever an IBC deposit named an allow-listed asset, without confirming the deposit arrived on the genuine Axelar channel. The upstream escrow contract checked a token's channel path and capped releases at what that channel had escrowed. When the fork was rebuilt to mint instead of escrow, those checks were removed and the new allow list was keyed by token name only, so the source was never authenticated. Because anyone can open an IBC channel to a contract, a counterfeit chain could deliver deposits the contract accepted as real. The gap dated to the repo's first commit in January 2023 and carried through a March 2026 migration, per Common Prefix and Secret's own report, and the fork was not re-audited. A second gap made the loss total: Axelar's escrow paid out the redemptions of the fake tokens, roughly 18 minutes after minting, with no anomaly alert or pause. Axelar's own code and IBC were not compromised.
Prevention analysis
Binding each allowed token to the one genuine Axelar channel would have rejected every deposit from the counterfeit chain, and nothing would have been minted.
Turning an escrow contract into a minting contract removed the checks that implicitly authenticated the source. An audit of that change would very likely have caught the missing check.
The escrow paid out its entire reserve to one wallet in minutes and nobody noticed for seven days. A per-channel outflow cap or alert would have stopped most of the drain.
Similar incidents
IBC tokens with nothing behind them were minted on a counterparty chain and redeemed against real collateral. Same lesson: an IBC asset is only as sound as the accounting on the chain that issues it.
A bridge minted wrapped assets for a deposit message it had not properly verified, then the unbacked tokens were redeemed for real funds. Same class: mint without authenticating where the message came from.
Remediation
Timeline
Get your protocol scored across 12 dimensions, or request ongoing coverage.