JUDAO's own sell-side transfer hook burned and mined JUDAO out of its PancakeSwap pair and re-synced the pair's reserve, and an attacker funded by a Moolah flash loan then swapped the JUDAO it had just sent to the pair for USDT, and an attacker funded by a Moolah flash loan called the pair's swap() directly against that gap to pull out USDT. DeFiHackLabs reports 205,259.49 USDT plus 36 BNB taken.
Security ·
Logged as “2026-04-28: Judao Sell Hook Reserve Drain on JUDAO”.
JUDAO, traded on BNB Smart Chain under the on-chain name T3 JUDAO, was a token with its own buy and sell logic built into _transfer rather than relying on a plain BEP-20 balance update. DeFiHackLabs's reconstruction, the only detailed account of this incident available, describes the token recognizing a transfer to its registered trading pair as a sell and running extra bookkeeping on that path: what its own analysis comment calls "sync(true) mining and isBurnPair reserve burns." The effect of that bookkeeping was to pull JUDAO out of the pair and re-sync its reserve downward, so the JUDAO the attacker then sent to the pair showed up as a large input against a shrunken reserve.
That matters because a PancakeSwap pair, built on the same model as Uniswap V2, does not ask where its tokens came from when it settles a swap() call. It only compares its current balance of each token to the reserve figure it cached after the last swap, and treats any increase beyond that cached figure as the input of the trade in progress. A normal fee-on-transfer token can already cause small mismatches of this kind; JUDAO's sell hook created a mismatch large enough to drain the pair outright, because the hook itself removed JUDAO from the pair and re-synced its reserve during the sell, leaving the JUDAO the attacker had just sent as a large apparent swap input.
The vulnerable contract is JUDAO itself, at 0xf55DFF7898930a2D28cDbC39D615b1624ac86888, confirmed against its registered basePair() by the proof of concept to be paired against USDT at 0x5D7b61e91cB59E90f7fAE8d0FE2e73976161592F. That specific pair carries essentially no liquidity today; dexscreener's current record of JUDAO's two active pairs shows a WBNB pair with about $283 in liquidity and a separate OSK pair with under a dollar, and the attacked USDT pair itself no longer appears among them, consistent with its reserves having been fully drained and never replenished.
The attack, by 0x5384B34C74024d6563B323351a4bBFA18432161B through attack contract 0x530904b5B5Ec86cca0528A682614F57F87e7f079, ran as one flash-loan-funded sequence:
swap() function directly with that gap as the input, at PancakeSwap's standard 0.25 percent fee, pulling USDT straight out of the pair.DeFiHackLabs's proof of concept puts the total at 205,259.49 USDT plus 36 BNB, consistent with the README's rounded 205K USDT plus 36 BNB. The attack transaction and the token's contract page have not been independently confirmed beyond DeFiHackLabs's reconstruction. Given JUDAO's thin, now-abandoned liquidity, no security firm or news outlet appears to have published an analysis of this incident, and no further detail is available from the DefimonAlerts post on Telegram that first flagged it.
The closest comparison in the site's other analyses is AIC, drained two months later through what its own entry describes as a "pair skim / reserve-mismatch exploit, flash-swap leveraged." Both incidents turn on the same underlying weakness in Uniswap V2-style pairs: a swap() call that trusts a balance increase over its cached reserve as real trade input, regardless of how that balance increase arrived. JUDAO's route to the mismatch was its own transfer hook; AIC reached the same gap through a flash swap instead, without a cooperating token.
What would have prevented it is either of two fixes: a trading pair should treat tokens as traded only if they arrived through the counterparty's transferFrom in the swap call itself, or a token like JUDAO should never alter a pair's balance through logic in _transfer that runs independently of an actual swap.
Query every incident yourself — free, no signup
Updated 9 Nov 2026. Figures are as reported by the sources listed above and are not independently verified.