Fake Swap Quotes Are Draining DeFi Wallets Through Uniswap Hooks

A Price You See Is Not Always the Price You Get

On September 14, liquidity aggregator 0x reported a sharp rise in malicious Uniswap v4 hooks engineered to display one swap quote and settle at a lower amount. In the worst cases 0x observed, traders walked away with 50% less than the figure shown on screen before they approved the transaction.

The attack works because aggregators are designed to find the most favorable route. A bad actor who controls a pool only needs to make it look attractive during the comparison phase – then execute something different at settlement.

What 0x Found When It Looked at the Hook Landscape

To size the problem, 0x analyzed 84,163 Uniswap v4 hooks across six chains using static analysis, dynamic analysis, and settled-trade observations. The dataset was labeled as of September 11. Of those hooks, 0x classified 19.4% as safe, 54.2% as malicious, and the remaining 26.4% as likely malicious – meaning fewer than one in five passed a clean bill of health.

One specific hook on the Base network, trading the ETH/NVDAc pair, recorded 6,516 fills in that window. Of those, 3,946 were charged fills. Fees on that hook ranged from 0% to 18%, with a median of 18% when a fee was actually applied. Total fees collected reached $143,037 as of September 11 – not a rounding error, and not a bug.

The mechanism behind hooks is straightforward enough on paper. Uniswap v4 hooks are optional external contracts that can execute before or after a swap, and selected permissions can adjust balance deltas. Uniswap’s own developer documentation explains this design, and the protocol explicitly warns users that independent third parties write hooks and that the code may be malicious or produce unintended consequences. The warning exists because the architecture was built with flexibility as a feature. Dynamic fees, custom accounting, and specialized liquidity logic all require this kind of extensibility. The same extensibility is also what allows a hook to recognize a quote request and return a different number at execution.

The practical impact on a user is easier to understand through comparison. Suppose an aggregator routes a swap through Pool A, which quotes 100 tokens, rather than Pool B, which quotes 98. Pool A wins the screen. If Pool A’s hook then settles for 80, the trader who trusted the better-looking number actually received less than Pool B would have delivered. The quote that looked worse was the safer trade.

This Pattern Has Been Building Since at Least March

The 0x report is not the first time this specific type of manipulation has surfaced. In July, routing infrastructure provider Enso documented what it called “toxic pools,” including a Polygon Uniswap v4 hook that used execution-environment signals to behave differently under simulation-like conditions, and an Ethereum Curve pool whose oracle behavior changed depending on how it was queried.

A March 0x study found a related pattern in proprietary liquidity on Base. Its propAMM analysis showed one market maker outperforming a reference automated market maker in 100% of sampled quote-time observations – but consistently settling worse at execution, typically by 5 to 10 basis points, or 0.05% to 0.10%. That gap looks small per trade. Across millions of transactions, the math compounds quickly.

Who Filters What, and What That Means for Traders

A September 8 scorecard from ClearTrace found zero or small median quote gaps for several sampled aggregators in Ethereum fork simulations. That test measured simulated quote accuracy, not the hook dataset 0x examined, so it does not contradict 0x’s findings – it just covers different terrain. The abuse appears concentrated in specific venues rather than spread evenly across swap routing infrastructure.

Routers are now the de facto gatekeepers. They decide which pools to simulate, which execution patterns to flag, and when an attractive-looking source should be excluded regardless of how good its quote appears. Wallets that route through these systems inherit whatever filtering decisions were made upstream – typically without any display of which liquidity was excluded or why. Users generally have no visibility into that layer.

Some infrastructure is responding by moving the comparison closer to settlement. KyberSwap said its Smart Settlement feature prepares multiple candidate pools for a swap hop, compares them on-chain at execution time, and atomically selects whichever candidate is offering the highest output at that moment. That approach makes quote manipulation harder to sustain because the winning pool is determined at the point of no return rather than during the simulation phase.

0x said it handled 81.92 million trades totaling $42.67 billion in volume during 2026 through September 14, with roughly 70% of those transactions touching Uniswap liquidity. That concentration means blocklist and vetting decisions made at the aggregator level determine which markets are practically reachable for a large share of DeFi users. 0x said it cuts off liquidity sources until execution issues are resolved, even when doing so makes its own displayed quotes look less competitive than rivals who have not yet excluded the same sources.

Protocol-level access stays open. App-level curation is where the real decisions are being made – and that hook on Base with $143,037 in collected fees was sitting inside a permissionless system the whole time.