kommunitas logo
BlogBlockchain

Kava Lend Exploit Explained: How a Dormant BUSD Market Drained $3.76 Million

Kava Lend Exploit Explained: How a Dormant BUSD Market Drained $3.76 Million

Kava Lend Exploit Explained: How a Dormant BUSD Market Drained $3.76 Million

Blockchain • November 18, 2026

By Priyo Harjiyono

TL;DR: On November 14, 2026, an attacker drained about 55.2 million KAVA, worth roughly $3.76 million at the pre-attack price, from Kava Lend, the lending app on the Kava blockchain, without stealing a single private key. The attack ran through a dormant market for BUSD, a stablecoin no longer minted since February 2023. By shrinking that market's net supply to two base units, the attacker made the protocol's interest math divide by almost nothing, inflating a BUSD deposit about 243,000-fold into roughly $1.25 billion of paper collateral. The attacker then borrowed KAVA 127 times, wiping each loan with a self-liquidation, bridged about 49.2 million KAVA to BNB Chain, and sold it for about $2.09 million. Kava's November 16 incident report confirmed 53.9 million KAVA taken, and the incident was not an infinite mint. As of November 18, no full technical post-mortem, frozen-funds total, or lender repayment plan has appeared.

Last updated: November 2026

Key Takeaways:

  • The bug was arithmetic, not access. One function divided interest by net supply and only refused when supply was exactly zero. Two base units passed the check.
  • A dead market did the damage. BUSD minting ended in February 2023, yet Kava Lend still counted BUSD deposits as collateral at a 50% loan-to-value.
  • Two flaws combined. The inflated index created phantom collateral, and a liquidation path that erased full debts let the attacker reuse the borrow cap 127 times.
  • The figures measure different things: 55.2 million KAVA borrowed ($3.76 million pre-attack), 53.9 million kept (Kava's confirmed figure), about $2.09 million realized from the sale.
  • Kava's November 16 report confirmed the loss, and a governance vote is set to shut HARD Lend's lending down permanently. Lender repayment and the 612,416 USDT blocked at Relay remain unresolved.

Most crypto hacks get described as stolen keys or broken smart contracts. The Kava Lend exploit was neither: every transaction the attacker sent was valid under the protocol's own rules. What broke was a division, guarded against exactly zero and nothing else. The attacker brought it two.

What Is Kava Lend?

Kava is a layer-1 blockchain that has run since 2019, built on the Cosmos SDK, and Kava Lend, now called HARD Lend, is its native money market, running as a Go chain module called x/hard. You deposit a coin into a shared pool to earn interest, and you borrow other coins by posting collateral worth more than the loan. A supply interest index tracks what the pool owes depositors: each time interest accrues, the index rises and every deposit grows by the same ratio. Against BUSD, a borrower could take out half the collateral's value. If a loan grows past its limit, anyone can liquidate it for a 2% reward.

By November 2026, Kava Lend was already being wound down: a governance proposal passed on September 28, 2026, ended incentives, capped borrowing, and told users to close their positions. Depositors left, but the BUSD market's collateral setting stayed on.

How the Exploit Worked, Step by Step

The division at the heart of it

Each time interest accrues, the module runs CalculateSupplyInterestFactor: it divides the new interest, minus the reserve cut, by net supply (cash plus borrows minus reserves), and multiplies the supply index by one plus that ratio:

totalSupply := cash.Add(borrows).Sub(reserves)
if totalSupply.IsZero() {
    return sdk.OneDec()
}
return (newInterest.Quo(totalSupply)).Add(sdk.OneDec())

Read the guard. If net supply is exactly zero, skip the division. If it is two base units, divide by two. A base unit is the smallest piece of a token the chain can count, so two base units of BUSD is $0.00000002. The same file's utilization function treats a tiny supply as a hazard. Only the interest-factor path got the zero-only check.

A market that was already empty

BUSD is the Binance-branded dollar stablecoin, and its issuer, Paxos, stopped minting it in February 2023. Just before the attack, Bitquery read the BUSD market as: 440,145 BUSD owed to lenders, 5,084 BUSD out on loan across 37 loans, 0.003 BUSD in cash, 7,618 BUSD of reserves, index at 2.76. Reserves count against supply, so net supply had drifted toward zero on its own.

Three minutes that inflated everything

Bitquery, The Block, and SigIntZero each reconstructed the attack from on-chain data, Kava's public API, and the source code:

  1. November 8 to 10. Repeated small BUSD deposit and borrow cycles.
  2. 04:01:12 UTC, November 14. The main wallet repaid all 37 open BUSD loans, then in one transaction deposited 3,801.60 BUSD, borrowed 1,267.20 BUSD, and withdrew 1,267.20 BUSD. Net supply: two base units.
  3. 04:01:18 UTC. Interest accrued over that supply of two. The factor multiplied by about 8,380 in a single block, from 2.762 to 23,147.
  4. 04:01 to 04:03 UTC. The factor climbed to 671,186.99, about 243,000 times its starting value. A 5,192.65 BUSD deposit now read about 1.26 billion BUSD.
  5. 04:03:58 UTC. The attacker repaid the BUSD loan and withdrew 7,618.45 BUSD, freezing the index. The inflated claim stayed on the books.

The borrow-and-liquidate loop

Kava Lend capped total KAVA borrowing at 448,358 KAVA, and about 447,640 KAVA was already borrowed. The attacker got around the cap with a loop between two wallets. Wallet A sent real BUSD to wallet B, which deposited it as collateral. A then used its inflated balance to withdraw the same BUSD and send it around again, leaving B with more collateral on paper each pass. B borrowed up to 445,719 KAVA per round, and in the next block A liquidated B. In a healthy market, liquidations are usually carried out by liquidation bots that race to close underwater loans for a small reward. Here the liquidator was the attacker's own wallet, and the auction lot was 0, 1, or 2 base units, because the BUSD backing B's deposit had already gone back to A. The code erased the full debt anyway, B kept the KAVA, and the borrow cap was empty again.

The loop ran 127 times over 98 minutes, from 04:04:21 to 05:42:04 UTC. SigIntZero's count of the helper wallet's borrows sums to 55,205,050.6 KAVA, matching Bitquery and The Block, and the module account holding the pooled KAVA fell from 55.33 million KAVA to a low of 0.19 KAVA, per The Block.

Timeline: From Rehearsal to Shutdown Vote

Date and time (UTC)Event
Sep 28, 2026Wind-down proposal passes. BUSD collateral factor stays at 50%.
Nov 8–10, 2026Small BUSD deposit and borrow cycles in the dormant market.
Nov 14, 04:01Attacker sets BUSD net supply to two base units.
Nov 14, 04:01–04:03BUSD supply index jumps about 243,000-fold, to 671,186.99.
Nov 14, 04:04–05:42127 borrow-and-liquidate rounds drain about 55.2 million KAVA.
Nov 14, 10:07–10:4248.4 million KAVA sold in 35 minutes. Price falls 36% during the sale.
Nov 14Upbit, Bithumb, and Coinone suspend KAVA deposits and withdrawals. KAVA falls about 40%.
Nov 16Kava's incident report confirms 53.9 million KAVA taken. Bithumb issues an investment warning.

Which Loss Figure Should You Trust?

The numbers measure different things:

FigureWhat it measuresSource
55.2M KAVA, about $3.76MGross KAVA borrowed, at the $0.068 pre-attack priceThe Block, Bitquery
About 53.9M KAVANet KAVA kept after repaying other borrowers' loansBitquery; confirmed by Kava
About $2.09MWhat the attacker realized selling into a falling marketThe Block, Bitquery
About $1.83MThe same 53.9M KAVA at the post-crash price near $0.034KuCoin, citing Kava's update

What depositors lost is a separate question. Kava's API still recorded about 55.21 million KAVA of deposits while the module account held about 1,987 KAVA as of November 16, per SigIntZero's API queries, and Kava has not said whether those claims will be made whole. DefiLlama classifies the incident as a token and share accounting failure caused by an arithmetic error.

Where the Money Went

Six hours passed between the first loan and the sale, and the price did not move in that window. At 09:59 UTC, about 49.2 million KAVA started moving through LayerZero to BNB Chain. Three minutes after the coins arrived, the selling began through MetaMask Swaps, an aggregator that splits trades across DEX pools. Over 35 minutes, the seller sold 48.4 million KAVA in 46 transfers, and the price fell 36% during the sale and 52% within two hours. The sale brought in about $2.09 million in BNB and USDT, per The Block, barely half the pre-attack value. That kind of one-way flow is also brutal for the liquidity providers in those pools: as each sale moved the pool ratio, their positions rebalanced into ever more KAVA, the same pool mechanics behind impermanent loss.

Within two hours, the proceeds had left BNB Chain by four routes, all through services that work without an account. The largest share went to Hyperliquid: about a third of the BNB went through Chainflip and LI.FI and arrived as USDC, where eight accounts received $1.17 million and each bought XMR1, a Monero-named Hyperliquid token, within a minute of its deposit, totaling 2,190 XMR1. Bitquery's reading is that those accounts belong to a Monero swap service, meaning the seller was most likely paid in Monero and the trail ends there; about 15% of the BNB went to Solana, where the trail also ends. The fourth route was blocked: the seller sent 612,416 USDT into the Relay bridge asking for ETH, and the ETH never arrived. Relay's record marks the order as failed with the reason "BLOCKED_WALLET," but also marks the deposit as withdrawable by the sender, so recovery depends on Kava or law enforcement asking Relay to hold it. As of November 18, no recovery has been confirmed.

What Kava and the Exchanges Did

Kava's first public word came at 11:28 UTC on November 14, more than seven hours after the first malicious transaction. About half an hour later, the team wrote: "To clarify: This is not an infinite mint. The event is isolated and contained within Kava Lend." At 14:05 UTC, Kava paused its Axelar bridge "out of an abundance of caution," although the stolen KAVA had left through LayerZero. The same day, Upbit, Bithumb, and Coinone suspended KAVA deposits and withdrawals, and Upbit put KAVA on its delisting watchlist, per CoinCentral's reporting. On November 16, Bithumb designated KAVA an investment-warning asset, with a decision expected between December 14 and 18, 2026.

On November 16, Kava published its incident report. It confirmed approximately 53.9 million KAVA taken from HARD Lend, identified a vulnerability in the lending application's accounting system, and said the exploit had been blocked and some attacker-controlled funds had been frozen. It said user deposits were "largely unaffected," and announced a governance proposal to permanently disable new lending and borrowing through a seven-day vote. The report did not explain how the BUSD collateral value was inflated, quantify how much had been frozen, give a US-dollar loss figure, or address whether depositors would be repaid. Those gaps are why independent reconstructions from Bitquery, The Block, and SigIntZero remain the primary technical record.

An independent reading of the module's parameters on November 16 found the minimum loan size set to $1 trillion: borrowing was off because no loan could ever be large enough. A governance proposal in voting would set every borrow cap to zero and restore the minimum to $10. The vulnerable function itself had not changed in the public repository. One independent analysis also reported a separate, unverified shortfall of about 23,492.86 ATOM in the ATOM market that appeared on November 10; Kava's report does not address it.

Why Dormant Markets Keep Breaking Lending Protocols

A lending market whose real liquidity has drained away is still a live set of formulas, and every ratio in it divides by a supply an attacker can set by hand. Kava Lend is the fifth known case of this shape: Hundred Finance (April 2023, about $7 million), Onyx (November 2023, about $2.1 million), Radiant (January 2024, about $4.5 million), and Sonne Finance (May 2024, about $20 million) all involved a near-empty market, a multiplier computed by dividing by what was in it, and an inflated position borrowing a real asset. The first four were new markets. Kava's was an old market on its way out.

On August 30, 2026, the Cronos lending protocol Tectonic lost an estimated $75 million after an attacker pumped its thinly traded TONIC token roughly 100 times in about 20 minutes and borrowed against the inflated valuation, per estimates from Weilin Li and PeckShield. Cronos validators halted the chain and rolled it back; only about $6 million reached Ethereum. Tectonic's contracts behaved as configured: the failure was a collateral factor that never asked whether the collateral could actually be sold. Kava's failure was arithmetic, but the shape is the same, and the Drift hack on Solana showed the failure does not have to be a stolen key: it drained at least $270 million through pre-signed admin transactions and a zero-timelock multisig.

Three lessons come out of the Kava case. Wind-downs need an order of operations: collateral factor to zero first, borrowing off second, and only then tell depositors to leave. Denominators need floors, not just zero checks: clamping how far an index can move in one block would have stopped the first block cold. And monitoring needs levers: SigIntZero's review found no pause path in the x/hard keeper code, so an alert could only start a human response, not stop the module.

What Lenders and KAVA Holders Should Check

If you hold KAVA in a wallet or on an exchange, the hack took none of your coins. It took about 40% of their price on the day, and the token traded near $0.034 on November 16. If you lent KAVA through Kava Lend, the picture is worse: the pool owes its lenders about 55.2 million KAVA and held almost none of it after the drain, and no repayment plan has been published as of November 18. If you have a position in HARD Lend, watch for Kava's promised guidance and the governance vote result, rather than assuming the pool will reopen normally.

Before you deposit into any lending protocol, the Kava case gives you a short checklist:

  1. Check for dead markets still counted as collateral. Look at each market's cash, borrows, and reserves, not just its total deposits: a market whose asset stopped being issued and whose cash is near zero has denominators an attacker can steer by hand.
  2. Check the collateral factor on deprecated assets. If a token's issuer has wound it down, the lending market should have been set to zero loan-to-value long ago.
  3. Check wind-down announcements. A protocol that has announced a shutdown deserves extra scrutiny: the dangerous window is after depositors leave and before the parameters are switched off.
  4. Check whether the protocol can pause, and read exchange notices after an incident. A protocol that can only be stopped by a governance vote has no emergency brake for a 98-minute drain, and warning designations like Bithumb's directly affect whether you can move or sell a token.

None of this is financial advice. The next drain of this kind will probably not be a new listing: it will be a market for a deprecated asset, in a protocol that announced a wind-down, with a collateral factor still above zero on the day it happens.

Frequently Asked Questions

Was the Kava Lend exploit an infinite mint?

No. The 55.2 million KAVA the attacker took was existing KAVA borrowed out of Kava Lend's pool, whose balance fell from 55.33 million KAVA to a low of 0.19 KAVA. What was inflated was an accounting number, the BUSD supply interest factor. No new KAVA was created.

How much did Kava Lend lose?

About 55.2 million KAVA was borrowed, worth roughly $3.76 million at the $0.068 pre-attack price. The attacker wallets kept about 53.9 million KAVA, the figure Kava confirmed. The attacker realized about $2.09 million selling on BNB Chain.

What was the root cause of the Kava Lend hack?

Two flaws in the x/hard module, by the reading of independent analysts. The supply interest factor divides by net supply and only checks for exactly zero, so a supply of two base units inflated the BUSD index about 243,000-fold. The liquidation path then cleared full debts while auctioning almost nothing, so the attacker could wipe each KAVA loan and reuse the borrow cap 127 times.

Has Kava published a post-mortem?

Not a full technical post-mortem. Kava's November 16 incident report confirmed about 53.9 million KAVA taken, the vulnerability blocked, some funds frozen, and a shutdown vote. It does not explain the mechanism, quantify frozen funds, or include a compensation plan. As of November 18, no further technical report has appeared.

Can KAVA lenders withdraw from Kava Lend?

Not in full, as far as the chain shows. Lenders are owed about 55.2 million KAVA, and the pool held almost none of it after the drain. No repayment plan had been published as of November 18.

Could other lending protocols have the same bug?

Yes. Any lending market that divides interest by net supply and only guards against zero is exposed when that market can be drained to a few base units. The same shape has hit Hundred Finance, Onyx, Radiant, and Sonne Finance since 2023. The defense is a floor under the denominator, a cap on how far an index can move in one block, and a zero collateral factor on deprecated assets.

Sources

Share This Article

This link will open in a new window