Morning Edition · Sunday, September 20, 2026Published at 1:45 AM EDT · New York
Prime Finance lost about 425 wrapped HYPE to a replayed but validly signed oracle report, Likwid lost 74.31 BNB to a missing reserve update, and MultiversX spent Saturday investigating an undisclosed mainnet issue.

Security researchers published three reproducible exploit demonstrations overnight, and none of them involved broken cryptography or a compiler trick. Each turned on a check a developer assumed someone else had made.
The largest, against the lending market Prime Finance, replayed real oracle data. The protocol's report-verification function was open to anyone and confirmed that a price report carried a valid signature from Chainlink's decentralized oracle network, but never confirmed the report was current. The attacker took a flash loan, submitted two previously signed reports, repriced the PRFI token from roughly $0.0021 to roughly $0.1101, bought the token from a thin pool, posted it as collateral at the inflated value and borrowed against it. About 425.5 wrapped HYPE left the reserve. Every signature involved was genuine.
The second, against the margin protocol Likwid on BNB Chain, cost 74.31 BNB. In the zero-leverage borrow path the contract updated three internal balances but not the one that feeds back into pool reserves, so the automated market maker quoted the same marginal price no matter how much value moved. The attacker repeated the cycle fourteen times. SlowMist flagged the loss. The third write-up shows a permissionless staking call inflating the reported net asset value of a Flamincome vault strategy, so withdrawals overpaid.
Separately, MultiversX said at 10:00 UTC it was investigating a potential mainnet issue and promised an update within twelve hours. The team disclosed no cause, and reported no confirmed losses. None of the three exploit write-ups records recovered funds or an attribution.
What this means
All three incidents sit outside the surface that audits and bug bounties normally price. A signature check that passes, a staking function that is meant to be open, and an accounting field that exists but goes unset are not code defects in the conventional sense, which is why reviewers scoped to contract logic keep missing them. Lenders and vault depositors carry the loss directly, and the practical cost falls to protocols that must now extend monitoring to freshness windows, privileged callers and every path where an external input is trusted because it is signed.
What to watch
Part of a tracked trend
Losses Move to Components That Worked as Designed
A growing share of DeFi losses will come not from buggy contract code but from components behaving exactly as specified — oracle forwarders, validator signature sets, governance votes and other trusted off-contract inputs — so audits and bug bounties scoped to on-chain code keep missing the failure surface, and protocols will be repeatedly forced to extend review, scope and monitoring to their privileged operational infrastructure.
Start a discussion in Townsquare.
More from this edition
Observations to monitor, not financial advice.
Synthesized from: DeFiHackLabs · DeFiHackLabs · DeFiHackLabs · crypto.news
Comments
0No comments yet.