Curve vs Uniswap V4: Liquidation Leaks and Hook Manipulation
Curve's LLAMMA algorithm manages crvUSD through continuous rebalancing, while Uniswap V4 hooks present risks where users can receive up to 50% less than quoted prices. These mechanisms create distinct profiles regarding rebalancing fee leakage and programmable price manipulation.
LLAMMA and crvUSD Liquidation
Curve manages crvUSD loans using the LLAMMA algorithm for soft liquidation. This mechanism spreads collateral across a range of price levels to adjust positions continuously. As the collateral price falls through these levels, the protocol sells portions of the deposit for crvUSD. If the price recovers, the protocol buys the collateral back. This process makes a drawdown less abrupt, but it does not prevent full liquidation during deep crashes. This rebalancing process also leaks value through trading fees and arbitrage. In March, Curve lowered average crvUSD borrow rates from 11.1% to 5.6%. The protocol also increased the crvUSD flashloan cap from 3M to 30M crvUSD. FastBridge enables crvUSD withdrawals from L2s like Arbitrum, Optimism, or Fraxtal to Ethereum in approximately 15 minutes using LayerZero and an Ethereum-side vault. Llamalend TVL reached $182M in August 2026, with supply at $89.2M and borrowing at $99M. crvUSD maintains a price within a tenth of a cent of $1. The GHO PegKeeper has a 3M crvUSD debt ceiling. Earlier vulnerabilities in Curve included a 2023 exploit caused by a Vyper compiler issue that allowed re-entrancy.
Uniswap V4 Hook Vulnerabilities
Malicious Uniswap v4 hooks quote one price and settle at another. Analysis of 84,163 hooks across six chains reveals significant vulnerabilities in the system. The 0x report classified 54.2% of these hooks as malicious and 26.4% as likely malicious. These hooks manipulate how aggregators pick routes by appearing more attractive during quote comparison. One Base hook trading ETH/NVDAc recorded 6,516 fills, 3,946 of which were charged. This hook collected $143,037 in fees by September 11, and the median fee for those charged fills reached 18%. In the most extreme trades observed by 0x, users received as much as 50% less at execution than the amount displayed in the quote, which creates a massive gap between the expected price and the actual settled amount for the trader. 0x routed $42.67 billion in volume through September 2026, and roughly 70% of that total volume touched Uniswap liquidity. UniswapX operates via a Dutch auction where third-party fillers compete and pay gas for the swapper. You must watch for these discrepancies if you rely on aggregator quotes. Can aggregators effectively filter these toxic pools before they hit the settlement layer?
| Feature | Curve crvUSD | Uniswap V4 Hooks |
|---|---|---|
| Primary Mechanism | LLAMMA soft liquidation | Programmable hooks |
| Risk Driver | Rebalancing fee leakage | Price/quote manipulation |
| Max Observed Loss | Continuous via arbitrage | 50% quote shortfall |
Comparison of Risk Models
Curve’s crvUSD system and Uniswap v4 present distinct risk profiles. Curve focuses on credit via crvUSD, while Uniswap V4 introduces programmable logic through hooks. LLAMMA rebalancing continuously moves collateral, which creates a constant drain on the borrower’s position through fees and arbitrage. Conversely, Uniswap v4 hooks can deceive traders by quoting a specific price and then settling at a significantly worse rate. This makes Uniswap’s programmable fee tiers a primary source of unexpected slippage. Uniswap v3 offered standard tiers of 0.05%, 0.30%, and 1%, while v4 allows creators to set any fee. The risk of Uniswap v4 hooks involves the manipulation of quotes, while the risk of Curve’s LLAMMA is the value lost during continuous rebalancing. Standard lending protocols use a single threshold for liquidation, but LLAMMA uses a range. This prevents the cliff effect of hard liquidation. In stablecoin pairs, providers often use a 0.01% tier to minimize slippage.
Join the discussion