# Qwen3.5

**Tool:** Qwen3.5 (Alibaba Cloud) **Type:** 3-round AI audit (Systematic → Economic → Triage) **Contracts reviewed:** evently.sol v5.4 · EventlyProfiles.sol v1.3 · EventlyMarketsV3.sol (LMSR, b=200) **Chain:** MegaETH (Chain ID 4326) — Solidity ^0.8.20 **Date:** March 2026 **Status:** Complete — v3.2 post-audit fixes verified

***

## Summary

| Severity  | Count |
| --------- | ----- |
| Critical  | 0     |
| High      | 0     |
| Medium    | 0     |
| Low       | 1     |
| Info      | 3     |
| **Total** | **4** |

**Overall Assessment:** The contract demonstrates strong security hygiene with proper CEI patterns, SafeERC20 usage, and correct LMSR implementation. Known fixes (F-02, A-03) are properly applied. No production blockers identified.

***

## Round 1 — Systematic Vulnerability Scan

### Reentrancy Protection

✅ **CEI Pattern:** All external-facing functions with state changes follow Checks-Effects-Interactions. State updates precede `safeTransfer`/`_safeTransferFrom` calls.

✅ **nonReentrant Modifier:** Applied to all functions with external calls: `buyShares`, `sellToAMM`, `createSellOrder`, `cancelSellOrder`, `disputeMarket`, `settleDispute`, `claimCreatorFees`, `redeemWinnings`, `claimCancelRefund`.

✅ **LOW-001 \[RESOLVED — L-01]:** `nonReentrant` has been added to `resolveMarket`, `cancelMarket`, and `slashMarket`. While CEI is followed (state changes before external calls), adding the modifier provides defense-in-depth against future refactoring risks.

```solidity
// Current: resolveMarket lacks nonReentrant
function resolveMarket(uint256 _marketId, uint256 _winningOption) external {
    // ... state changes ...
    _cancelAllOrders(_marketId); // External transfers via _safeTransferFrom
}

// Recommended:
function resolveMarket(uint256 _marketId, uint256 _winningOption) external nonReentrant {
    // ...
}
```

**Recommended Fix:** Add `nonReentrant` to `resolveMarket`, `finalizeMarket`, `cancelMarket`, and `slashMarket`.

### Access Control

✅ Clear separation between `onlyAdmin`, `onlyWhitelisted`, and creator-only functions. No privilege escalation vectors.

✅ Creator collateral logic correctly uses `creatorCollateralReturned` flag to prevent double-claims.

✅ Invite code system properly prevents reuse via `inviteCodeUsed` mapping.

### Integer Arithmetic & Fee Precision

✅ Solidity ^0.8.20 provides automatic overflow revert.

✅ Fee rounding dust absorbed by `resolverCut`:

```solidity
uint256 resolverCut = totalFee - creatorCut - treasuryCut; // Absorbs rounding error
```

✅ `setFees` validates `totalBps <= MAX_TOTAL_FEE_BPS (500)`.

### Token Handling

✅ All USDm transfers use `safeTransfer`/`safeTransferFrom`.

✅ Proper use of `_mint`, `_burn`, `_safeTransferFrom` with correct token ID encoding (`marketId * MAX_OPTIONS + optionIndex`).

✅ P2P sell orders properly escrow shares to contract before listing.

### LMSR Mathematical Correctness

✅ `_lmsrCost` correctly implements `C(q) = b * ln(Σ exp(qᵢ/b))` with proper fixed-point conversion via PRBMath.

✅ `quoteBuy` uses tolerance `1e15` (0.001 shares) with proper bracketing and capacity validation.

✅ `getPrice` maintains `Σ Pᵢ = 1e18` (100%) via normalized exponentials.

⚠️ **INFO-001:** Upper bound `maxQ = 26,000e18` yields `q/b ≈ 130`, approaching the PRBMath `exp()` overflow threshold (\~133). Currently safe, but future library changes or parameter adjustments could trigger overflow.

**Recommended Mitigation:** Add explicit check: `require(q[i] * 1e18 / b_ < 133e18, "LMSR overflow risk")`.

***

## Round 2 — Economic Attack Analysis

### LMSR Solvency Guarantee

✅ **Invariant Verified:** `poolBalance + subsidyDeposited ≥ total winning shares × 1e18`

The LMSR mechanism ensures solvency because:

1. Initial subsidy = `b * ln(n)` covers worst-case uniform→concentrated transition
2. Trader payments (net of fees) increase `poolBalance` proportionally to shares minted
3. Fees are deducted *before* LMSR cost calculation, so shares minted correspond to net collateral
4. Redemption logic draws from `poolBalance` first, then `subsidyDeposited`, with explicit exhaustion check

### Fee Accounting Integrity

✅ Default 2.5% split (1% creator, 1% treasury, 0.5% resolver) is correctly enforced.

✅ Creator fees accrue in `creatorAccruedFees[_marketId]` and are only claimable post-finalization with `creatorFeePaid` guard.

✅ Admin market creator fees correctly redirected to treasury when `isAdminMarket || creator == address(0)`.

### Order Book Security

✅ `MAX_ORDERS_PER_BOOK = 200` (A-03 fix) prevents order book flooding.

✅ P2P orders filled before AMM; P2P trades don't affect LMSR quantities, preventing price oracle manipulation.

✅ Cheapest-first sorting with proper slippage protection via `_minShares`/`_minUsdm`.

⚠️ **INFO-002:** Binary search tolerance `1e15` in `quoteBuy` may cause rounding at the `MIN_TRADE` boundary. A trade of exactly `MIN_TRADE` could yield 0 or `2e15` shares due to tolerance bounds.

**Impact:** Negligible for practical trade sizes; protected by `require(ammShares >= MIN_TRADE)`.

### Resolution MEV Analysis

✅ `resolveMarket` requires `block.timestamp >= bettingDeadline`, preventing pre-deadline resolution. Status changes to `Resolved` before order cancellation, blocking further trades.

✅ 24-hour `DISPUTE_WINDOW` with collateral requirement prevents spam disputes.

### Subsidy Exhaustion Attack

✅ `redeemWinnings` checks `fromSubsidy <= m.subsidyDeposited` before drawing from subsidy. LMSR solvency guarantee ensures this holds under normal operation.

### claimCancelRefund Sequential Call Safety

✅ F-02 fix snapshots `optionSupply` before burning, ensuring fair pro-rata distribution across sequential claimants.

***

## Round 3 — Adversarial Triage

### Fund Theft Vectors

No theft vectors found. Funds only exit via: `redeemWinnings` (requires burning winning shares), `claimCancelRefund` (requires burning shares in cancelled market), `sellToAMM` (requires burning shares with AMM pricing), or authorized fee withdrawals.

### Permanent Bricking Vectors

No permanent bricking path found. Admin key compromise allows fee changes and treasury withdrawals, but cannot freeze user redemptions. `claimCancelRefund` remains callable regardless of admin status.

**Centralization Risk:** Admin key compromise enables fee increases to 5% cap, resolver pool redirection, market slashing, and treasury withdrawal. Recommended mitigation: deploy admin behind multisig or timelock.

### Edge Cases

✅ Zero-volume market: subsidy covers theoretical payouts; no actual shares to redeem.

✅ Single bettor monopoly: LMSR cost escalation and `maxQ` cap prevent infinite share acquisition.

✅ All bets on losing side: unclaimed `poolBalance` remains in contract (design choice, not vulnerability). No sweep mechanism.

⚠️ **INFO-003:** `closeBetting` function lacks event emission, reducing off-chain observability of market state transitions.

**Recommended Fix:** Add `emit BettingClosed(_marketId);`.

***

## Findings Summary Table

| ID       | Severity | Contract             | Title                                              | Status                                                    |
| -------- | -------- | -------------------- | -------------------------------------------------- | --------------------------------------------------------- |
| LOW-001  | Low      | EventlyMarketsV3.sol | Missing `nonReentrant` on resolution functions     | Acknowledged                                              |
| INFO-001 | Info     | EventlyMarketsV3.sol | LMSR upper bound near `exp()` overflow threshold   | Monitor                                                   |
| INFO-002 | Info     | EventlyMarketsV3.sol | Binary search tolerance edge case near `MIN_TRADE` | Acknowledged                                              |
| INFO-003 | Info     | EventlyMarketsV3.sol | Missing event in `closeBetting` function           | **Resolved — v3.2: BettingClosed event added (QWEN-I03)** |

***

## Production Blockers

✅ **No Critical/High findings blocking deployment.** The contract is suitable for production with the following recommendations:

1. Add `nonReentrant` to resolution functions (LOW-001) as defense-in-depth
2. Deploy admin behind multisig/timelock to mitigate centralization risk
3. Monitor LMSR parameter adjustments to maintain `q/b < 133` safety margin

***

## Severity Scale

| Severity | Impact                                         | Description                                                |
| -------- | ---------------------------------------------- | ---------------------------------------------------------- |
| Critical | Fund loss, permanent bricking, protocol halt   | Exploitable vulnerability causing irreversible damage      |
| High     | Significant fund risk, major logic bypass      | Exploitable under realistic conditions with notable impact |
| Medium   | Limited fund risk, partial logic bypass        | Exploitable but with constraints or limited impact         |
| Low      | No fund risk, best practice violation          | Code quality, gas optimization, or defense-in-depth issues |
| Info     | Observability, documentation, minor edge cases | Suggestions for improvement without security impact        |

*Audit conducted via static analysis, formal reasoning of LMSR invariants, and adversarial simulation. Dynamic testing recommended pre-deployment.*