1inch Bug Bounty Report: Aqua Edition, H1 2026

Finding 1 — Inverted Scale Formula in Concentrated Liquidity

Researcher: nathan47 | Severity: 🟡 Medium

A sign error can be a goldmine for attackers.

Deep inside Aqua's concentrated liquidity engine, nathan47 found exactly that: the scale update formula was inverted. Instead of tightening the price range after a swap, the formula moved it in the opposite direction, steadily breaking the AMM's core invariant with every trade. Over time, the drift created a systematic arbitrage window against maker positions. The bug was hiding in plain sight behind correct-looking code.

🔧 Fix: swap-vm PR #82 «[PT1-396] fixed concentrate» (XYCConcentrate rewritten in stateless-model); additional changes — #105, #113, #114

________________________________________________________________________________________________________________________________________________

Findings 2 & 3 — Dutch Auction Decay Bug: uint32 Truncation and Non-Functional Balance Instructions

Researchers: Bz & Xmanuel | Severity: 🔵 Low (×2)

Dutch auctions are elegant: the price starts high and decays over time until a taker fills the order. It's a classic mechanism for efficient price discovery. But it just... didn't work.

Xmanuel discovered that decayFactor was declared as a uint32 - but the calculation expected a 1e18-scaled fixed-point number. The truncation was silent and total. Every decay curve became a flat line.

Bz went deeper and found that both balance instructions for Dutch auctions - _dutchAuctionBalanceIn1D and _dutchAuctionBalanceOut1D - were non-functional. Makers who thought they were posting time-weighted price discovery orders were actually posting static limit orders, exposed to adverse selection for the order's entire lifetime.

Two separate researchers. Two separate angles. One broken feature - fully exposed.

🔧 Fix commit: https://github.com/1inch/swap-vm/pull/25/changes

________________________________________________________________________________________________________________________________________________

Finding 4 — Strategy Hash Collision Triggering Global Reentrancy Lock

Researcher: mayoo0x0 | Severity: 🔵 Low

Reentrancy locks exist to prevent one of DeFi's oldest attack vectors. But what happens when two completely different makers end up with the same strategy hash?

mayoo0x0 traced a scenario where a hash collision between distinct makers' strategies would trigger the global reentrancy lock — effectively blocking atomic routing for Aqua Apps entirely. No funds stolen directly. Just a hard stop, mid-execution. In a system where atomicity is the guarantee, a DoS like this could be a serious reliability failure.

🔧 Awareness commit: 13434dc (25 Feb) / PR #85 «added doc about orderHash collisions»

________________________________________________________________________________________________________________________________________________

Finding 5 — MakerTraits Hook Flag Mismatch Enabling Fund Redirection

Researcher: Z3rco | Severity: 🟠 High

Z3rco identified a mismatch in how MakerTraits hook flags are encoded versus how they're interpreted during execution. By carefully crafting the hook encoding, a malicious taker could manipulate transaction execution - redirecting or outright stealing maker funds during an otherwise normal-looking swap. With a working proof of concept demonstrating real fund loss, this earned the program's only High-severity payout of H1.

It's a reminder that in a programmable swap engine, the boundary between "configuration" and "code" is thin - and every serialization boundary is a potential attack surface.

🔧 Fix commit: https://github.com/1inch/sdks/pull/32

________________________________________________________________________________________________________________________________________________

Finding 6 — TWAP Unit Mismatch Allowing Full Stream Drain

Researcher: dantehrani | Severity: 🔵 Low

TWAP orders exist to pace liquidity over time - to sell gradually rather than all at once. The mechanism relies on a cap: you can't drain more than your allotted share in a given window.

dantehrani found that the cap was comparing amountIn against available amountOut. Two different units, two different tokens, being compared directly. The result: a determined taker could drain the entire TWAP stream in a single transaction, bypassing the rate-limiting entirely.

🔧 Fix commit: https://github.com/1inch/swap-vm/pull/105

Note: The issue was technically out of scope, but the researcher still received a bounty.

________________________________________________________________________________________________________________________________________________

Findings 7 & 8 — Storage Slot Offset Bug and Gas Unit Mismatch in Fee Machinery

Researcher: Jugger63 | Severity: 🔵 Low (x2)

Jugger63 closed out the paid findings with a sharp pair of low-level catches in Aqua's fee and decay machinery.

First: the Decay._decayXD function had an offset direction bug. Decay state was being written to one storage slot and read from a completely different one. Every decay calculation was operating on stale or zero data, silently corrupting strategy behavior over time - the kind of bug that's nearly invisible in unit tests but damaging in production.

Second: the gas compensation calculation in BaseFeeAdjuster's ExactIn path was mixing incompatible units - gas units against token amounts. The adjustment was orders of magnitude off, making the entire gas compensation mechanism effectively dead code.

🔧 Fix commit: https://github.com/1inch/swap-vm/pull/70