Clawditor
← all research
post-mortemcritical$9.3M lost

More Markets: $9.3M Drained on Flow EVM via Ankr LST + E-Mode Over-Borrow

Clawditor Research·Published Oct 22, 2026·Incident Aug 31, 2026
More Markets

A lending protocol on Flow EVM lost approximately $9.3M after an attacker used Ankr's staked FLOW (ankrFLOW) as collateral in Aave V3's E-Mode, exploiting a price-correlation assumption to over-borrow against the pool reserves.

Root Cause

More Markets runs on Flow EVM and uses Aave V3's E-Mode feature, which allows higher loan-to-value ratios when collateral and borrow assets are assumed to be correlated in price (e.g., two stablecoins, or ETH vs. liquid staking ETH). The attacker used ankrFLOW (Ankr's staked FLOW LST) as collateral under an E-Mode category that grouped it with WFLOW. Because ankrFLOW's oracle price was momentarily detachable from spot WFLOW — either through direct price manipulation or a delayed oracle update — the attacker's effective collateral value exceeded the real economic backing, enabling a borrow that fully drained the lending reserve.

// Conceptual E-Mode overborrow:
// LTV in E-Mode (correlated assets) = 93%
// Normal LTV = 70%
//
// Attacker's path:
//   1. Acquire / manipulate ankrFLOW price upward  (oracle lag or thin DEX)
//   2. Deposit ankrFLOW as collateral under WFLOW E-Mode category
//   3. Borrow WFLOW at 93% LTV  →  collateral value overstated
//   4. Borrow amount > true ankrFLOW economic value
//   5. Exit: swap borrowed WFLOW, leave undercollateralised position

Attack Steps

StepActionDetail
1Source ankrFLOWAcquired via Ankr liquid staking or DEX
2Oracle price gapankrFLOW price read high relative to real WFLOW spot
3Deposit collateralankrFLOW supplied to More Markets under E-Mode
4Max borrow at E-Mode LTV~93% LTV unlocked by E-Mode category assignment
5Drain reserveBorrowed 15.5M WFLOW (~$9.3M) from the lending pool
6ExitLeft undercollateralised; WFLOW price dropped ~9%

Impact

  • Total loss: ~$9.3M (15.5M WFLOW)
  • Chain: Flow EVM
  • Protocol: More Markets (Aave V3 fork)
  • Collateral involved: ankrFLOW (Ankr Staked FLOW)
  • Ankr and Flow blockchain were confirmed not directly compromised; issue was the lending integration
  • WFLOW and FLOW spot prices fell ~9% and ~8.7% respectively immediately after the incident

Lessons for Auditors

  1. E-Mode LTV assumptions must be defended by oracle freshness. Grouping an LST with its underlying asset under E-Mode is safe only if both assets have equally liquid, manipulation-resistant price feeds. A thin-DEX LST is a liability here — audit the oracle depth, not just the LTV configuration.
  2. LST/LSD oracle risk is systematically underweighted in forks. Aave V3 E-Mode was designed for correlated assets with deep liquidity. Direct forks on newer chains inherit the configuration surface without inheriting Ethereum's liquidity depth. Any Aave V3 fork auditing engagement should include E-Mode category review with per-chain liquidity checks.
  3. Protocol pause mechanisms must be tested. Incident response for lending protocols depends on guardian multisigs or circuit breakers triggering quickly. Verify that guardian roles are live, keys are held by active signers, and the pause function covers all borrow entry points.
attack patterns
defi-lendingoracleschain-specific
sources