Dev.to Security πŸ” Cybersecurity πŸ‘ 0 πŸ“– 8 min read

B18TimeBank Exploit: Signed Vouchers Drain 404,540 USDC

Originally published at B18TimeBank Exploit: Signed Vouchers Drain 404,540 USDC On 17 November 2026 at 18:50:59 UTC, an address on Base called withdrawLevelReward on B18TimeBank's settlement contract and received 404,54

Originally published at B18TimeBank Exploit: Signed Vouchers Drain 404,540 USDC

On 17 November 2026 at 18:50:59 UTC, an address on Base called withdrawLevelReward on B18TimeBank's settlement contract and received 404,540.468304 USDC in the same transaction. Blockaid disclosed the incident at 19:51:29 UTC, said an attacker had redeemed signed level-reward vouchers and withdrawn the USDC from that contract, and reported about $404,540 drained across one transaction. Independently, we checked the same Base transaction and the settlement contract's USDC balance: it fell from 511,325.678867 USDC to 106,785.210563 USDC, a drop of exactly 404,540.468304 USDC.

B18TimeBank has not published an official loss figure or a public statement. This is a defensive account of a public incident, not a reconstruction of how to repeat it, and it does not establish that the remaining funds are safe.

What Happened

B18TimeBank presents itself as a cycle-based, or "time bank," protocol. B18's documentation describes users depositing into time-structured products and receiving settlement at maturity, including accrued value. We have not seen a public post-mortem that maps this incident onto a specific product, and we are not inferring one from the docs.

Blockaid labeled four on-chain objects, all of which appear in the transaction it cited:

  • Exploiter: 0x28189EaeB578569E74FC1548b87440a384a51973
  • Example exploit transaction: 0x4b3061fe…db4966
  • Victim settlement: 0x654520c9bAE2D8369043547bf0574C9F036CD054
  • Victim pool: 0xC95106353EC7272202677A2a79378071888B0a1F

The transaction is on Base, in block 58755418, at 18:50:59 UTC. The exploiter called the settlement contract directly with selector 0x8bb5a6b2, which 4byte.directory records as withdrawLevelReward(uint256). The single argument decodes as 790,000 Γ— 10^18, matching the 790,000 B18 that moved in the same transaction. The call was successful and paid 404,540.468304 USDC, Circle's USDC on Base (0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913), from the settlement contract to the exploiter.

That dollar figure is the token amount itself, not a later estimate of sale proceeds. The Defiant and subsequent write-ups cited the same transfer on BaseScan, which matches our check.

What the Transaction Moved

Before the call, the settlement contract held 511,325.678867 USDC and the pool held zero USDC. After it, the settlement contract held 106,785.210563 USDC and the pool still held zero. The exploiter's USDC balance rose from 3.721043 to 404,544.189347, an increase of the same 404,540.468304 USDC. The pool's balance of the B18 Protocol token at 0xA0c8b0C06Ff087d168B79177bab402F8844825Cb fell from 1,040,191.99 B18 to 408,191.99 B18, a net outflow of 632,000 B18.

Inside the transaction the B18 movements were larger than that net figure, and they netted out on the settlement contract. 790,000 B18 first moved from the pool to the settlement contract. The settlement contract then sent the same 790,000 B18 back out in four pieces: 395,000 B18 (50%) to 0xCc9bBEAA5Fc06E29594A596968B3Dabe1EB10a9B, 158,000 B18 (20%) to the burn address 0x000…dEaD, 158,000 B18 (20%) back to the pool, and 79,000 B18 (10%) to 0xb485d8A0167599976ed2738469b3AdB2fc8080ee. The settlement contract's own B18 balance was unchanged. We have not seen an official description of those four recipients, and we do not assign roles to them.

The 404,540.468304 USDC payout came from the settlement contract, not the pool. The 106,785 USDC that remained after the call is the figure later reports checked at about 19:53 UTC. It is not a second payout, and these two contracts are not a measure of protocol-wide deposits.

Signed Rewards, and What Is Still Unknown

Blockaid's initial alert said the attacker redeemed signed level-reward vouchers before withdrawing USDC. A signed voucher is an off-chain authorization: a trusted signer attests that a reward may be claimed, and a contract is supposed to pay only if that attestation is valid and still unused. The first alert did not explain how the attacker obtained vouchers the contract would honor β€” a stolen or mis-used signer key, a replay of a valid voucher, a missing binding to the rightful beneficiary, or a bug in verification would all be consistent with the same public description, and none of those causes has been established.

We have not found verified source for the settlement contract on Sourcify, and we are not reverse-engineering the bytecode to fill that gap. Until a public post-mortem appears, the verified facts stop at the function that was called, the balances that changed, and Blockaid's description of the path.

The pattern is still worth a defensive reminder. A signed reward is only as safe as the checks around it. The sketch below is illustrative, not B18 source, and is not a reproduction of the incident. It only states properties a voucher redemption should not be able to skip:

// Simplified, illustrative invariant β€” not B18 source.
// A signed reward must bind beneficiary, amount, a unique nonce and an expiry.
function signedRewardIsBound(
    address beneficiary,
    uint256 amount,
    uint256 nonce,
    uint256 expiry,
    bytes calldata signature
) view {
    require(block.timestamp <= expiry, "expired");
    require(!used[nonce], "already used");
    bytes32 digest = keccak256(abi.encode(beneficiary, amount, nonce, expiry));
    require(recoverSigner(digest, signature) == trustedSigner, "bad signature");
    require(msg.sender == beneficiary, "wrong claimant");
}

Authorization bugs in settlement contracts are not new. In the Pennysia Settlement approval drain, a flexible settlement function could be steered into spending other users' token approvals. In the Ankr bridge forged-receipts incident, an authorization that was supposed to represent a real event was accepted when it should not have been. The surfaces differ; the lesson does not: a contract that pays out against a signature or a receipt has to reject reused, unbound, or unexpected claims. The same caution applies to settlement accounting, a point we also made in the TranchePool UTC rollover settlement exploit. A signed voucher should not be a way around what the product actually owes.

Timeline

Times below are in UTC on 17 November 2026.

  • 18:50:59: The Blockaid exploiter called withdrawLevelReward on the settlement contract in 0x4b3061fe…db4966. 790,000 B18 moved from the pool through the settlement contract, and 404,540.468304 USDC was paid to the exploiter. The settlement USDC balance fell from 511,325.68 to 106,785.21.
  • 18:59:22–19:23:03: The exploiter started moving USDC onward. 100,010 USDC went to 0x2F8145Efd9cE4D6E683f432ac47B5E211c987323 in two transfers, 304,533.19 USDC went to 0x8461324c21c4b6ebf6C566da73e442850c4c3d3e in three transfers β€” including 104,533.19 USDC at 19:23:03, the later transfer highlighted in contemporaneous reports β€” and 1 USDC went to a third address. Only dust remained on the exploiter in a later check.
  • 19:51:29: Blockaid reported an ongoing exploit, about $404,540 USDC drained from the settlement contract in one transaction, and published the addresses above. A follow-up post listed the exploiter, the example transaction, the settlement contract, and the pool.
  • 19:53 approximately: Subsequent reporting recorded about 106,785 USDC still in the settlement contract and about 408,192 B18 still in the pool, with zero USDC in the pool β€” matching the post-transaction balances we measured.

Aftermath

At the time of writing, B18TimeBank had not posted a public response. Its most recent post on that account was from 6 July 2026, well before this incident. We also found no confirmation from B18 or from Base that the payout path had been paused or otherwise contained. Blockaid described the exploit as ongoing in its 19:51 UTC alert.

The settlement contract still held 106,781.505931 USDC when we checked it after the first reports. The small decrease from 106,785.210563 USDC came from later transfers to other addresses; we do not assign those to the same attacker, and we do not know whether the original path remained usable.

CertiK's public project page for B18 lists a 23 March 2026 audit of B18Settlement.sol and B18Token.sol. That listing is not a statement about this incident.

This article is technical commentary, not financial advice.

Lessons for Builders

Treat a signed reward as a payment instruction, not as a suggestion. The contract has to recover a trusted signer, bind the beneficiary and the amount, enforce an expiry, and mark a nonce or voucher id as used before any funds move. If any of those checks is missing, a signature becomes a cheque that can be cashed more than once, or by the wrong person.

Keep the authorized amount inside the product's accounting. A level reward or maturity payout should not be able to pay more than the product owes, even if a signature presents a larger figure. Simulate the claim against actual reserves and against the user's recorded entitlement.

Do not let a redemption drain a shared reserve without a circuit breaker. This withdrawal emptied most of the settlement contract's USDC in one call and left more than 106,000 USDC behind. A cap per voucher, a pause switch, and monitoring on large stablecoin outflows from settlement contracts are the controls that still matter after the fact.

Publish addresses, then follow through on containment. Blockaid's pointers made the transaction easy to verify. A pause, a signer rotation, or a public "do not interact" notice is what depositors need next, and none of those had appeared from the project at the time of writing.

What Users Should Know

If you deposited into B18TimeBank or hold B18, follow updates only from the project's own channels once they speak, beginning with B18TimeBank on X and B18's documentation site. Treat unsolicited "refund" or "recovery" messages as phishing until the project describes a process.

The 106,781 USDC still in the settlement contract is not a promise that remaining deposits are safe, and it is not a figure for total user exposure. Whether the same withdrawal path still works has not been publicly confirmed. Do not deposit additional funds until the team has published a fix and said that the settlement contracts are safe to use. Revoking any approvals you granted to the settlement contract or the pool cannot undo a payout that has already left, but it can limit later misuse of an allowance.

Sources

Closing Thoughts

The B18TimeBank loss did not require a new token to be minted or a failure of Base itself. A settlement contract paid 404,540 USDC on a withdrawLevelReward call that Blockaid described as the redemption of signed level-reward vouchers, and most of that USDC had already been moved onward before the public alert. If you are building a protocol that settles rewards, maturities, or other authorized claims against a shared reserve, talk to our team or check our audit pricing to have voucher validation, signer controls and settlement accounting reviewed before a single call is the first production test of those checks.

πŸ“° Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β€” full credit and traffic to the original publisher.