The Invariant Off by a Hundred

Medium

· Backend

Incident

SEV-1

BUG-URANIUM-0428

Priority: Critical · reported by protocol monitoring

Time to fix: 20 min

Failing checks on the shipped build

  • a draining swap is rejected

    a tiny input cannot take 90% of the pool: expected an error to be raised

How to work this incident

  1. Read the report: write down the expected behaviour and the actual one in a sentence each.
  2. Reproduce it: run the visible tests and find the failing check that matches the report.
  3. Read the logs and the failing checks for the first value that is wrong — the earliest wrong value is nearer the cause than the loudest error.
  4. Follow the request through the handler, the validation and the service it calls; find the first step whose output disagrees with the report.
  5. Fix the cause rather than the symptom, run the visible tests again, and submit against the hidden ones — then write the root cause down.

Inspired by

Uranium Finance · 2021

Modelled on the Uranium Finance exploit (April 2021): Uranium, a fork of Uniswap v2, changed the scale of its swap-fee arithmetic from 1,000 to 10,000, but the constant-product check at the end of swap still compared against the reserves scaled by 1,000². The invariant was therefore about 100 times weaker than intended, and attackers drained the pools by asking for far more output than their input paid for — tens of millions of dollars' worth.

This project is a reconstruction. fees.py defines the fee scale; pair.py scales balances by it and then checks the product against the old constant.

Fix swap so the invariant uses the same scale as the balances.

Code map

fees = bug_require("./fees.py")


def swap(reserve0, reserve1, amount0_in, amount1_out):
    """Token0 in, token1 out. Returns the new [reserve0, reserve1]."""
    if amount0_in <= 0 or amount1_out <= 0:
        raise ValueError("INSUFFICIENT_AMOUNT")
    if amount1_out >= reserve1:
        raise ValueError("INSUFFICIENT_LIQUIDITY")
    balance0 = reserve0 + amount0_in
    balance1 = reserve1 - amount1_out
    adjusted0 = balance0 * fees.SCALE - amount0_in * fees.FEE
    adjusted1 = balance1 * fees.SCALE
    if adjusted0 * adjusted1 < reserve0 * reserve1 * 1000 ** 2:
        raise ValueError("K")
    return [balance0, balance1]

Sign in to edit the file, run the tests and submit a fix

A free account — no card, no trial clock. You come straight back here.

The Invariant Off by a Hundred — Python Bug Hunt

Modelled on the Uranium Finance exploit (April 2021): Uranium, a fork of Uniswap v2, changed the scale of its swap-fee arithmetic from 1,000 to 10,000, but…

  • Language: Python
  • Layer: Backend
  • Difficulty: Medium
  • Concepts: Money, Validation
  • Modelled on: Uranium Finance · 2021
  • Visible tests: a swap at the quoted price is accepted; a draining swap is rejected
  • Reward: 50 XP for a complete fix

Incident

  • Severity: SEV-1 (priority: Critical)
  • Ticket: BUG-URANIUM-0428
  • Reported by: protocol monitoring
  • Time to fix: 20 minutes, the target for a medium incident

Failing checks on the shipped build

  • a draining swap is rejected — a tiny input cannot take 90% of the pool: expected an error to be raised

How to work this incident

  1. Read the report: write down the expected behaviour and the actual one in a sentence each.
  2. Reproduce it: run the visible tests and find the failing check that matches the report.
  3. Read the logs and the failing checks for the first value that is wrong — the earliest wrong value is nearer the cause than the loudest error.
  4. Follow the request through the handler, the validation and the service it calls; find the first step whose output disagrees with the report.
  5. Fix the cause rather than the symptom, run the visible tests again, and submit against the hidden ones — then write the root cause down.

Briefing

Modelled on the Uranium Finance exploit (April 2021): Uranium, a fork of Uniswap v2, changed the scale of its swap-fee arithmetic from 1,000 to 10,000, but the constant-product check at the end of swap still compared against the reserves scaled by 1,000². The invariant was therefore about 100 times weaker than intended, and attackers drained the pools by asking for far more output than their input paid for — tens of millions of dollars' worth.

This project is a reconstruction. fees.py defines the fee scale; pair.py scales balances by it and then checks the product against the old constant.

Fix swap so the invariant uses the same scale as the balances.

Bug report

BUG-URANIUM-0428 · Priority: Critical · Reported by: protocol monitoring

swap(reserve0, reserve1, amount0_in, amount1_out) (token0 in, token1 out):

  • raise ValueError if amount0_in <= 0 or amount1_out <= 0
  • raise ValueError if amount1_out >= reserve1
  • balance0 = reserve0 + amount0_in, balance1 = reserve1 - amount1_out
  • adjusted0 = balance0 SCALE - amount0_in FEE, adjusted1 = balance1 * SCALE
  • raise ValueError("K") unless adjusted0 adjusted1 >= reserve0 reserve1 * SCALE**2
  • otherwise return [balance0, balance1]

With this rule, fees.get_amount_out(...) is exactly the largest accepted output: the quote passes, the quote + 1 is rejected.

Observed: a 1,000-unit input withdraws 900,000 of a 1,000,000 reserve.

Logs

[pair 0x8a..] swap in0=1000 out1=900000 k_check=pass
[pair 0x8a..] reserve1 1000000 -> 100000

The code as shipped

src/amm/pair.py (editable)

fees = bug_require("./fees.py")


def swap(reserve0, reserve1, amount0_in, amount1_out):
    """Token0 in, token1 out. Returns the new [reserve0, reserve1]."""
    if amount0_in <= 0 or amount1_out <= 0:
        raise ValueError("INSUFFICIENT_AMOUNT")
    if amount1_out >= reserve1:
        raise ValueError("INSUFFICIENT_LIQUIDITY")
    balance0 = reserve0 + amount0_in
    balance1 = reserve1 - amount1_out
    adjusted0 = balance0 * fees.SCALE - amount0_in * fees.FEE
    adjusted1 = balance1 * fees.SCALE
    if adjusted0 * adjusted1 < reserve0 * reserve1 * 1000 ** 2:
        raise ValueError("K")
    return [balance0, balance1]

Read-only context: src/amm/fees.py.

Open the hunt to start the incident clock, edit the files, run the visible tests and submit against the hidden ones, then write the root cause down. More Python bug hunts.