Overview

Code4rena (C4) is an open organization consisting of security researchers, auditors, developers, and individuals with domain expertise in smart contracts.

A C4 code contest is an event in which community participants, referred to as Wardens, review, audit, or analyze smart contract logic in exchange for a bounty provided by sponsoring projects.

During the code contest outlined in this document, C4 conducted an analysis of the NFTX smart contract system written in Solidity. The code contest took place between December 16—December 22 2021.

Wardens

27 Wardens contributed reports to the NFTX contest:

1. cccz
2. hyh
3. leastwood
4. WatchPug (jtp and ming)
5. cmichel
6. robee
7. GreyArt (hickuphh3 and itsmeSTYJ)
8. pauliax
9. gzeon
10. Ruhum
11. ych18
12. csanuragjain
13. pedroais
14. jayjonah8
15. sirhashalot
16. 0x1f8b
17. PPrieditis
18. 0x0x0x
19. p4st13r4 (0x69e8 and 0xb4bb4)
20. defsec
21. Dravee
22. shenwilly
23. BouSalman
24. saian

This contest was judged by LSDan.

Final report assembled by liveactionllama and itsmetechjay.

Summary

The C4 analysis yielded an aggregated total of 58 unique vulnerabilities and 124 total findings. All of the issues presented here are linked back to their original finding.

Of these vulnerabilities, 3 received a risk rating in the category of HIGH severity, 17 received a risk rating in the category of MEDIUM severity, and 38 received a risk rating in the category of LOW severity.

C4 analysis also identified 19 non-critical recommendations and 47 gas optimizations.

Scope

The code under review can be found within the C4 NFTX contest repository, and is composed of 11 smart contracts written in the Solidity programming language and includes 3072 lines of Solidity code.

Severity Criteria

C4 assesses the severity of disclosed vulnerabilities according to a methodology based on OWASP standards.

Vulnerabilities are divided into three primary risk categories: high, medium, and low.

High-level considerations for vulnerabilities span the following key areas when conducting assessments:

• Malicious Input Handling
• Escalation of privileges
• Arithmetic
• Gas use

Further information regarding the severity criteria referenced throughout the submission review process, please refer to the documentation provided on the C4 website.

High Risk Findings (3)

[H-01] buyAndSwap1155WETH() function may cause loss of user assets

Submitted by cccz

In the NFTXMarketplaceZap.sol contract, the buyAndSwap1155WETH function uses the WETH provided by the user to exchange VaultToken, but when executing the _buyVaultToken method, msg.value is used instead of maxWethIn. Since msg.value is 0, the call will fail.

function buyAndSwap1155WETH(
uint256 vaultId,
uint256[] memory idsIn,
uint256[] memory amounts,
uint256[] memory specificIds,
uint256 maxWethIn,
) public payable nonReentrant {
require(idsIn.length != 0);
uint256 count;
for (uint256 i = 0; i <idsIn.length; i++) {
uint256 amount = amounts[i];
require(amount> 0, "Transferring <1");
count += amount;
}
INFTXVault vault = INFTXVault(nftxFactory.vault(vaultId));
uint256 redeemFees = (vault.targetSwapFee() * specificIds.length) + (
vault.randomSwapFee() * (count-specificIds.length)
);
uint256[] memory swapAmounts = _buyVaultToken(address(vault), redeemFees, msg.value, path);

In extreme cases, when the user provides both ETH and WETH (the user approves the contract WETH in advance and calls the buyAndSwap1155WETH function instead of the buyAndSwap1155 function by mistake), the _buyVaultToken function will execute successfully, but because the buyAndSwap1155WETH function will not convert ETH to WETH, The user’s ETH will be locked in the contract, causing loss of user assets.

function _buyVaultToken(
uint256 minTokenOut,
uint256 maxWethIn,
) internal returns (uint256[] memory) {
uint256[] memory amounts = sushiRouter.swapTokensForExactTokens(
minTokenOut,
maxWethIn,
path,
block.timestamp
);

return amounts;
}
  - uint256[] memory swapAmounts = _buyVaultToken(address(vault), redeemFees, msg.value, path);
+ uint256[] memory swapAmounts = _buyVaultToken(address(vault), redeemFees, maxWethIn, path);

0xKiwi (NFTX) confirmed and resolved

[H-02] The return value of the _sendForReceiver function is not set, causing the receiver to receive more fees

Submitted by cccz, also found by WatchPug

In the NFTXSimpleFeeDistributor.sol contract, the distribute function is used to distribute the fee, and the distribute function judges whether the fee is sent successfully according to the return value of the _sendForReceiver function.

function distribute(uint256 vaultId) external override virtual nonReentrant {

if (distributionPaused || allocTotal == 0) {
return;
}

uint256 leftover;
for (uint256 i = 0; i <length; i++) {
uint256 amountToSend = leftover + ((tokenBalance * _feeReceiver.allocPoint) / allocTotal);
amountToSend = amountToSend> currentTokenBalance? currentTokenBalance: amountToSend;
if (!complete) {
leftover = amountToSend;
} else {
leftover = 0;
}
}

In the _sendForReceiver function, when _receiver is not a contract, no value is returned. By default, this will return false. This will make the distribute function think that the fee sending has failed, and will send more fees next time.

function _sendForReceiver(FeeReceiver memory _receiver, uint256 _vaultId, address _vault, uint256 amountToSend) internal virtual returns (bool) {
// If the receive is not properly processed, send it to the treasury instead.

// If the allowance has not been spent, it means we can pass it forward to next.
} else {
}
}

Proof of Concept

https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXSimpleFeeDistributor.sol#L157-L168

https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXSimpleFeeDistributor.sol#L49-L67

function _sendForReceiver(FeeReceiver memory _receiver, uint256 _vaultId, address _vault, uint256 amountToSend) internal virtual returns (bool) {
// If the receive is not properly processed, send it to the treasury instead.

// If the allowance has not been spent, it means we can pass it forward to next.
} else {
}
}

0xKiwi (NFTX) confirmed, but disagreed with high severity and commented:

Good catch, thank you. Disagreeing with severity though since this is a permissioned contract, no user funds are at risk and this would most likely cause some failures.

We aren’t using any EOAs as receivers in production or testing, so this has not been caught. Thank you.

0xKiwi (NFTX) resolved

LSDan (judge) commented:

I agree with the warden on this one. Funds are directly at risk and the likelihood of this occurring is 100%. I’m not sure if it matters if the funds are user funds or protocol funds. This would eventually have become a big problem that affected the protocol’s ability to function.

[H-03] A vault can be locked from MarketplaceZap and StakingZap

Submitted by p4st13r4, also found by cmichel, GreyArt, hyh, jayjonah8, leastwood, pauliax, shenwilly, and WatchPug

Any user that owns a vToken of a particular vault can lock the functionalities of NFTXMarketplaceZap.sol and NFTXStakingZap.sol for everyone.

Every operation performed by the marketplace, that deals with vToken minting, performs this check:

require(balance == IERC20Upgradeable(vault).balanceOf(address(this)), "Did not receive expected balance");

A malicious user could transfer any amount > 0 of a vault’vToken to the marketplace (or staking) zap contracts, thus making the vault functionality unavailable for every user on the marketplace

Proof of Concept

https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXMarketplaceZap.sol#L421

https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXMarketplaceZap.sol#L421

Remove this logic from the marketplace and staking zap contracts, and add it to the vaults (if necessary)

0xKiwi (NFTX) confirmed, but disagreed with high severity and commented:

Valid concern, confirmed. And disagreeing with severity.

0xKiwi (NFTX) resolved

LSDan (judge) commented:

In this case I agree with the warden’s severity. The attack would cause user funds to be locked and is incredibly easy to perform.

Medium Risk Findings (17)

[M-01] Missing non reentrancy modifier

Submitted by robee

The following functions are missing reentrancy modifier although some other pulbic/external functions does use reentrancy modifer. Even though I did not find a way to exploit it, it seems like those functions should have the nonReentrant modifier as the other functions have it as well..

  NFTXMarketplaceZap.sol, receive is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, __SimpleFeeDistributor__init__ is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, initializeVaultReceivers is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, changeReceiverAlloc is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, removeReceiver is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, setTreasuryAddress is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, setLPStakingAddress is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, setInventoryStakingAddress is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, setNFTXVaultFactory is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, pauseFeeDistribution is missing a reentrancy modifier
NFTXSimpleFeeDistributor.sol, rescueTokens is missing a reentrancy modifier
NFTXStakingZap.sol, setLPLockTime is missing a reentrancy modifier
NFTXStakingZap.sol, setInventoryLockTime is missing a reentrancy modifier
NFTXStakingZap.sol, provideInventory721 is missing a reentrancy modifier
NFTXStakingZap.sol, provideInventory1155 is missing a reentrancy modifier
NFTXStakingZap.sol, addLiquidity721ETH is missing a reentrancy modifier
NFTXStakingZap.sol, addLiquidity1155ETH is missing a reentrancy modifier
NFTXStakingZap.sol, addLiquidity721 is missing a reentrancy modifier
NFTXStakingZap.sol, addLiquidity1155 is missing a reentrancy modifier
NFTXStakingZap.sol, receive is missing a reentrancy modifier
NFTXStakingZap.sol, rescue is missing a reentrancy modifier
NFTXV1Buyout.sol, emergencyWithdraw is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, __NFTXVault_init is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, finalizeVault is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, setVaultFeatures is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, assignDefaultFeatures is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, setFees is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, disableVaultFees is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, deployEligibilityStorage is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, setManager is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, mint is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, redeem is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, swap is missing a reentrancy modifier
NFTXVaultUpgradeable.sol, flashLoan is missing a reentrancy modifier
PalmNFTXStakingZap.sol, setLockTime is missing a reentrancy modifier
PalmNFTXStakingZap.sol, addLiquidity721 is missing a reentrancy modifier
PalmNFTXStakingZap.sol, addLiquidity1155 is missing a reentrancy modifier
PalmNFTXStakingZap.sol, receive is missing a reentrancy modifier

0xKiwi (NFTX) disputed

LSDan (judge) increased severity to medium and commented:

I’m updating this [from a low] to a medium. Reentrancy represents a real and significant risk (as evident by ETC existing) and should be protected against regardless of if you can foresee the external event that causes lack of protection to be an issue.

2 — Med (M): vulns have a risk of 2 and are considered “Medium” severity when assets are not at direct risk, but the function of the protocol or its availability could be impacted, or leak value with a hypothetical attack path with stated assumptions, but external requirements.

Submitted by GreyArt

The addReceiver() function fails to check if the _receiver already exists. This could lead to the same receiver being added multiple times, which results in erroneous fee distributions.

The receiver would receive more than expected (until the duplicate entry has been removed).

Have a mapping address => bool isReceiver that will update whenever receivers are added, modified to a new address or removed.

0xKiwi (NFTX) acknowledged, but disagreed with medium severity and commented:

Valid concern, but this is a permissioned function.

LSDan (judge) commented:

I think this one is much more likely. I would suggest adding a check because the problem is easy to create and much harder to notice.

[M-03] NFTXMarketplaceZap.sol#buyAnd***() should return unused weth/eth back to msg.sender instead of to

Submitted by WatchPug

https://github.com/code-423n4/2021-12-nftx/blob/194073f750b7e2c9a886ece34b6382b4f1355f36/nftx-protocol-v2/contracts/solidity/NFTXMarketplaceZap.sol#L226-L249

function buyAndSwap721WETH(
uint256 vaultId,
uint256[] memory idsIn,
uint256[] memory specificIds,
uint256 maxWethIn,
) public nonReentrant {
require(idsIn.length != 0);
INFTXVault vault = INFTXVault(nftxFactory.vault(vaultId));
uint256 redeemFees = (vault.targetSwapFee() * specificIds.length) + (
vault.randomSwapFee() * (idsIn.length - specificIds.length)
);
_swap721(vaultId, idsIn, specificIds, to);

emit Swap(idsIn.length, amounts[0], to);

// Return extras.
WETH.transfer(to, remaining);
}

For example:

If Alice calls buyAndSwap721WETH() to buy some ERC721 and send to Bob, for slippage control, Alice put 1000 ETH as maxWethIn, the actual cost should be lower.

Let’s say the actual cost is 900 ETH.

Expected Results: Alice spend only for the amount of the actual cost (900 ETH).

Actual Results: Alice spent 1000 ETH.

0xKiwi (NFTX) acknowledged, but disagreed with medium severity and commented:

I think the idea in this is that if a contract is buying for someone else, the zap handles the refund instead of the contract originating the purchase. But this is a valid concern, thank you

LSDan (judge) commented:

This does result in a loss of funds if the user sends the wrong amount. I agree with the warden’s severity rating.

[M-04] NFTXStakingZap and NFTXMarketplaceZap’s transferFromERC721 transfer Cryptokitties to the wrong address

Submitted by hyh

transferFromERC721(address assetAddr, uint256 tokenId, address to) should transfer from msg.sender to to. It transfers to address(this) instead when ERC721 is Cryptokitties. As there is no additional logic for this case it seems to be a mistake that leads to wrong NFT accounting after such a transfer as NFT will be missed in the vault (which is to).

Proof of Concept

NFTXStakingZap: transferFromERC721 https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXStakingZap.sol#L416

NFTXMarketplaceZap: transferFromERC721 https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXMarketplaceZap.sol#L556

Both functions are called by user facing Marketplace buy/sell and Staking addLiquidity/provideInventory functions.

Now:

  // Cryptokitties.
data = abi.encodeWithSignature("transferFrom(address,address,uint256)", msg.sender, address(this), tokenId);

To be:

  // Cryptokitties.
data = abi.encodeWithSignature("transferFrom(address,address,uint256)", msg.sender, to, tokenId);

0xKiwi (NFTX) confirmed, but disagreed with medium severity and commented:

This was intentional, as I thought it was needed for the contract to require custody, but it should work fine to send directly to the vault.

0xKiwi (NFTX) resolved

[M-05] Pool Manager can frontrun fees to 100% and use it to steal the value from users

Submitted by pedroais

Pool Manager can front-run entry fee to 100% and users could lose all their deposits.

Proof of Concept

Considering:
The pool manager is the creator of the pool.
Anyone can create a pool.
Manager is not a trusted actor.

Anyone can create a pool and get people to join. If there is a big deposit admin could front-run the transaction and set the fee to max which is uint(1 ether) = 10**18 (100% as this is a per token fee).

The manager could benefit from this by having other pool assets deposited in staking so he would receive fees in Vtokens and could use them to withdraw the nfts.

Add a timelock to change fees. In that way, frontrunning wouldn’t be possible and users would know the fees they are agreeing with.

0xKiwi (NFTX) acknowledged, but disagreed with high severity and commented:

Most users aren’t on vaults that aren’t finalized. We warn users for any vaults that arent finalized and we don’t present them on our website. Acknowledging and disagreeing with severity.

LSDan (judge) decreased severity to medium and commented:

In my view, this is a medium risk. While user funds are at direct risk, the likelihood of this happening or being worth the effort is low. As the sponsor states, it’s very rare for a user to interact with an un-finalized vault. The user would have to be directly linked to the vault and then ignore the giant warning presented front and center in the UI. If that warning were to be removed, however, the risk would increase. This external requirement is the only reason I’m going with medium and not low.

[M-06] xToken Approvals Allow Spenders To Spend More Tokens

Submitted by leastwood

The approve function has not been overridden and therefore uses xToken shares instead of the equivalent rebalanced amount, i.e. the underlying vault token amount.

Proof of Concept

The approved spender may spend more tokens than desired. In fact, the approved amount that can be transferred keeps growing as rewards continue to be distributed to the XTokenUpgradeable contract.

Many contracts also use the same amount for the approve call as for the amount they want to have transferred in a subsequent transferFrom call, and in this case, they approve an amount that is too large (as the approved shares amount yields a higher rebalanced amount).

The _allowances field should track the rebalanced amounts (i.e. the equivalent vault token amount) such that the approval value does not grow.

The transferFrom needs to be overridden and approvals should then be subtracted by the transferred vault token amount, not shares.

0xKiwi (NFTX) acknowledged, but disagreed with high severity and commented:

Not sure if I agree with this severity. If I approve for xTokens, I’m using xTokens, not the underlying token.

LSDan (judge) decreased severity to medium and commented:

This is a medium risk, not high. External assumptions (malicious contracts) are requires for any attack regarding the approval being too high.

[M-07] Rewards can be stolen

Submitted by cmichel

The NFTXInventoryStaking contract distributes new rewards to all previous stakers when the owner calls the receiveRewards function. This allows an attacker to frontrun this receiveRewards transaction when they see it in the mem pool with a deposit function. The attacker will receive the rewards pro-rata to their deposits. The deposit will be locked for 2 seconds only (DEFAULT_LOCKTIME) after which the depositor can withdraw their initial deposit & the rewards again for a profit.

The rewards can be gamed this way and one does not actually have to stake, only be in the staking contract at the time of reward distribution for 2 seconds. The rest of the time they can be used for other purposes.

Distribute the rewards equally over time to the stakers instead of in a single chunk on each receiveRewards call. This is more of a “streaming rewards” approach.

0xKiwi (NFTX) confirmed and commented:

Thanks for the report.

This is unfortunately unavoidable but streaming isn’t a bad idea. Will consider. Thank you.

Confirming.

[M-08] Low-level call return value not checked

Submitted by cmichel

The NFTXStakingZap.addLiquidity721ETHTo function performs a low-level .call in payable(to).call{value: msg.value-amountEth} but does not check the return value if the call succeeded.

Impact

If the call fails, the refunds did not succeed and the caller will lose all refunds of msg.value - amountEth.

Revert the entire transaction if the refund call fails by checking that the success return value of the payable(to).call(...) returns true.

0xKiwi (NFTX) confirmed and commented:

Nice catch, thank you. Confirmed.

0xKiwi (NFTX) resolved

[M-09] Bypass zap timelock

Submitted by gzeon

The default value of inventoryLockTime in NFTXStakingZap is 7 days while DEFAULT_LOCKTIME in NFTXInventoryStaking is 2 ms. These timelock value are used in NFTXInventoryStaking to eventually call _timelockMint in XTokenUpgradeable.

function _timelockMint(address account, uint256 amount, uint256 timelockLength) internal virtual {
uint256 timelockFinish = block.timestamp + timelockLength;
timelock[account] = timelockFinish;
emit Timelocked(account, timelockFinish);
_mint(account, amount);
}

The applicable timelock is calculated by block.timestamp + timelockLength, even when the existing timelock is further in the future. Therefore, one can reduce their long (e.g. 7 days) timelock to 2 ms calling deposit in NFTXInventoryStaking

Proof of Concept

function _timelockMint(address account, uint256 amount, uint256 timelockLength) internal virtual {
uint256 timelockFinish = block.timestamp + timelockLength;
if(timelockFinish > timelock[account]){
timelock[account] = timelockFinish;
emit Timelocked(account, timelockFinish);
}
_mint(account, amount);
}

0xKiwi (NFTX) disputed

0xKiwi (NFTX) confirmed and commented:

After taking another look, this is definitely accurate. Thank you!

0xKiwi (NFTX) resolved

Submitted by hyh

Double spending of fees being distributed will happen in favor of the first fee receivers in the feeReceivers list at the expense of the last ones. As _sendForReceiver doesn’t return success for completed transfer when receiver isn’t a contract, the corresponding fee amount is sent out twice, to the current and to the next fee receiver in the list. This will lead to double payments for those receivers who happen to be next in the line right after EOAs, and missed payments for the receivers positioned closer to the end of the list as the funds available are going to be already depleted when their turn comes.

Proof of Concept

distribute use _sendForReceiver to transfer current vault balance across feeReceivers: https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXSimpleFeeDistributor.sol#L67

_sendForReceiver returns a boolean that is used to move current distribution amount to the next receiver when last transfer failed. When _receiver.isContract is false nothing is returned, while safeTransfer is done: https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXSimpleFeeDistributor.sol#L168

This way _sendForReceiver will indicate that transfer is failed and leftover amount to be added to the next transfer, i.e. the amountToSend will be spent twice: https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXSimpleFeeDistributor.sol#L64

Now:

function _sendForReceiver(FeeReceiver memory _receiver, uint256 _vaultId, address _vault, uint256 amountToSend) internal virtual returns (bool) {
...
} else {
}
}

To be:

function _sendForReceiver(FeeReceiver memory _receiver, uint256 _vaultId, address _vault, uint256 amountToSend) internal virtual returns (bool) {
...
} else {
return true;
}
}

0xKiwi (NFTX) confirmed and resolved

[M-11] NFTXVaultFactoryUpgradeable implementation can be replaced in production breaking the system

Submitted by hyh

NFTXVaultFactory contract holds information regarding vaults, assets and permissions (vaults, _vaultsForAsset and excludedFromFees mappings). As there is no mechanics present that transfers this information to another implementation, the switch of nftxVaultFactory to another address performed while in production will break the system.

Proof of Concept

setNFTXVaultFactory function allows an owner to reset nftxVaultFactory without restrictions in the following contracts:

Either restrict the ability to change the factory implementation to pre-production stages or make nftxVaultFactory immutable by allowing changing it only once:

Now:

function setNFTXVaultFactory(address newFactory) external virtual override onlyOwner {
nftxVaultFactory = INFTXVaultFactory(newFactory);
}

To be:

function setNFTXVaultFactory(address newFactory) external virtual override onlyOwner {
require(nftxVaultFactory == address(0), "nftxVaultFactory is immutable");
nftxVaultFactory = INFTXVaultFactory(newFactory);
}

If the implementation upgrades in production is desired, the factory data migration logic should be implemented and then used atomically together with the implementation switch in all affected contracts.

0xKiwi (NFTX) confirmed and commented:

This is not a contract that is designed to be replaced, but upgraded. But it is a valid concern that these assistant contracts can have their factory be changed and rendered broken. (even if it were permissioned)
Confirming.

0xKiwi (NFTX) resolved

[M-12] buyAndSwap1155WETH Does Not Work As Intended

Submitted by leastwood

The buyAndSwap1155WETH function in NFTXMarketplaceZap aims to facilitate buying and swapping ERC1155 tokens within a single transaction. The function expects to transfer WETH tokens from the msg.sender account and use these tokens in purchasing vault tokens. However, the _buyVaultToken call in buyAndSwap1155WETH actually uses msg.value and not maxWethIn. As a result, the function will not work unless the user supplies both WETH and native ETH amounts, equivalent to the maxWethIn amount.

Proof of Concept

https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXMarketplaceZap.sol#L284-L314

function buyAndSwap1155WETH(
uint256 vaultId,
uint256[] memory idsIn,
uint256[] memory amounts,
uint256[] memory specificIds,
uint256 maxWethIn,
) public payable nonReentrant {
require(idsIn.length != 0);
uint256 count;
for (uint256 i = 0; i < idsIn.length; i++) {
uint256 amount = amounts[i];
require(amount > 0, "Transferring < 1");
count += amount;
}
INFTXVault vault = INFTXVault(nftxFactory.vault(vaultId));
uint256 redeemFees = (vault.targetSwapFee() * specificIds.length) + (
vault.randomSwapFee() * (count - specificIds.length)
);
_swap1155(vaultId, idsIn, amounts, specificIds, to);

emit Swap(count, swapAmounts[0], to);

// Return extras.
WETH.transfer(to, remaining);
}

Tools Used

Manual code review. Discussions with Kiwi.

Consider updating the buyAndSwap1155WETH function such that the following line of code is used instead of this.

uint256[] memory swapAmounts = _buyVaultToken(address(vault), redeemFees, maxWethIn, path);

0xKiwi (NFTX) confirmed and resolved

[M-13] Dishonest Stakers Can Siphon Rewards From xToken Holders Through The deposit Function In NFTXInventoryStaking

Submitted by leastwood

xTokens is intended to be a representation of staked vault tokens. As the protocol’s vaults accrue fees from users, these fees are intended to be distributed to users in an inconsistent fashion. NFTXInventoryStaking is one of the ways users can stake vault tokens. Deposits are timelocked for 2 seconds by default, essentially rendering flash loan attacks redundant. However, it is more than likely that the same user could withdraw their xToken deposit in the next block (assuming an average block time of just over 13 seconds).

Hence, if a well-funded attacker sees a transaction to distribute rewards to xToken holders, they could deposit a large sum of vault tokens and receive a majority share of the rewards before withdrawing their tokens in the following block. Additionally, the attacker can force distribute rewards in NFTXSimpleFeeDistributor as there is no access control on the distribute function.

This issue allows users to siphon user’s rewards from the protocol, intended to be distributed to honest vault token stakers.

Proof of Concept

Consider the following exploit scenario:

• Currently there are 1000 shares and 1000 base tokens in the XTokenUpgradeable contract.
• Honest actor, Alice, calls distribute in NFTXSimpleFeeDistributor which attempts to send 200 base tokens as rewards for xToken holders accrued via protocol usage.
• Bob sees a transaction to reward xToken holders and frontruns this transaction by staking vault tokens, minting 1000 shares and 1000 base tokens.
• Rewards are distributed such that XTokenUpgradeable has 2000 shares and 2200 base tokens.
• Bob unstakes his tokens and exits the pool, redeeming his 1000 shares for 1100 base tokens.
• As a result, Bob was able to siphon off 100 base tokens without having to stake their tokens for the same time period that Alice had staked her tokens for.
• This unfair distribution can be abused again and again to essentially reward dishonest actors over honest staking participants such as Alice.

Tools Used

Manual code review. Discussions with Kiwi.

Consider adding a delay to users token deposits into the XTokenUpgradeable such that miners cannot feasibly censor a transaction for the specified time interval and users cannot frontrun a transaction to distribute rewards. The interval should be chosen such that enough time is provided for the transaction to be included in a block, given poor network conditions.

I.e. If the chosen interval is 20 blocks. Miners must be able to censor the rewards distribution for 20 blocks. This is unlikely as there would need to be sufficient miner collusion for value to be extracted from the protocol. Additionally, an interval of 20 blocks means that stakers who attempt to enter the pool upon seeing the transaction in the mempool won’t be rewarded for such behaviour.

It is also essential that the distribute function in NFTXSimpleFeeDistributor is restricted to a given role, ensuring malicious users cannot control at what point rewards are distributed.

Alternatively, PoolTogether has a Time-Weighted-Average-Balance (TWAB) implementation which can be used as reference. This would ensure the fairest distribution of rewards to stakers, however, there are additional gas costs associated with this implementation. Hence, unless the protocol intends to be used primarily on L2 protocols, this solution should be avoided.

0xKiwi (NFTX) acknowledged, but disagreed with high severity and commented:

While this attack is possible, without available flash liquidity, this attack vector requires a lot of (possibly difficult to acquire) capital to execute. Disagreeing with severity and acknowledging.

LSDan (judge) decreased severity to medium and commented:

I agree with the sponsor that the risk of this happening is almost zero. Yes it’s technically possible but the funds lost are going to be minimal and the attacker will almost definitely pay more in slippage and gas fees than they make. That said, this is a direct attack which results in a loss of user funds so making it less than medium risk seems disingenuous.

[M-14] Return variable can remain unassigned in _sendForReceiver

Submitted by sirhashalot, also found by pauliax

The _sendForReceiver() function only sets a return function in the “if” code block, not the “else” case. If the “else” case is true, no value is returned. The result of this oversight is that the _sendForReceiver() function called from the distribute() function could sucessfully enter its else block if a receiver has isContract set to False and successfully transfer the amountToSend value. The ditribute() function will then have leftover > 0 and send currentTokenBalance to the treasury. This issue is partially due to Solidity using implicit returns, so if no bool value is explicitly returned, the default bool value of False will be returned.

This problem currently occurs for any receiver with isContract set to False. The _addReceiver function allows for isContract to be set to False, so such a condition should not result in tokens being sent to the treasury as though it was an emergency scenario.

Proof of Concept

The else block is missing a return value https://github.com/code-423n4/2021-12-nftx/blob/194073f750b7e2c9a886ece34b6382b4f1355f36/nftx-protocol-v2/contracts/solidity/NFTXSimpleFeeDistributor.sol#L167-L169

Tools Used

VS Code “Solidity Visual Developer” extension

Verify that functions with a return value do actually return a value in all cases. Adding the line return true; can be added to the end of the else block as one way to resolve this.

Alternatively, if isContract should never be set to False, the code should be designed to prevent a receiver from being added with this value.

0xKiwi (NFTX) confirmed and resolved

[M-15] No access control on assignFees() function in NFTXVaultFactoryUpgradeable contract

Submitted by ych18

If the Vault owner decides to set factoryMintFee and factoryRandomRedeemFee to zero, any user could call the function NFTXVaultFactoryUpgradeable.assignFees() and hence all the fees are updated.

0xKiwi (NFTX) confirmed and commented:

This function is left over from some upgrades. It will be removed. Thank you.

0xKiwi (NFTX) resolved

[M-16] Malicious receiver can make distribute function denial of service

Submitted by cccz

In the NFTXSimpleFeeDistributor.sol contract, the distribute function calls the _sendForReceiver function to distribute the fee

function distribute(uint256 vaultId) external override virtual nonReentrant {

if (distributionPaused || allocTotal == 0) {
return;
}

uint256 leftover;
for (uint256 i = 0; i <length; i++) {
uint256 amountToSend = leftover + ((tokenBalance * _feeReceiver.allocPoint) / allocTotal);
amountToSend = amountToSend> currentTokenBalance? currentTokenBalance: amountToSend;
if (!complete) {
leftover = amountToSend;
} else {
leftover = 0;
}
}

function _sendForReceiver(FeeReceiver memory _receiver, uint256 _vaultId, address _vault, uint256 amountToSend) internal virtual returns (bool) {
// If the receive is not properly processed, send it to the treasury instead.

// If the allowance has not been spent, it means we can pass it forward to next.
} else {
}
}

Proof of Concept

https://github.com/code-423n4/2021-12-nftx/blob/main/nftx-protocol-v2/contracts/solidity/NFTXSimpleFeeDistributor.sol#L157-L166

The contract can store the fee sent to the receiver in a state variable, and then the receiver can take it out by calling a function.

0xKiwi (NFTX) confirmed, but disagreed with high severity and commented:

This is a permissioned entity, but this is a valid concern. User funds not at risk and the dao can remove the malciious receiver (if it ever gets there in the first place). Good thinking.

0xKiwi (NFTX) resolved

LSDan (judge) decreased severity to medium and commented:

This is a medium risk, not high. The attack has external requirements and is relatively easy for the DAO to mitigate.

[M-17] transfer return value is ignored

Submitted by robee, also found by 0x1f8b, cmichel, csanuragjain, defsec, hyh, leastwood, sirhashalot, and WatchPug

Need to use safeTransfer instead of transfer. As there are popular tokens, such as USDT that transfer/transferFrom method doesn’t return anything. The transfer return value has to be checked (as there are some other tokens that returns false instead revert), that means you must

1. Check the transfer return value

Another popular possibility is to add a whiteList. Those are the appearances (solidity file, line number, actual line):

NFTXStakingZap.sol, 401, IERC20Upgradeable(vault).transfer(to, minTokenIn-amountToken);
PalmNFTXStakingZap.sol, 195, pairedToken.transfer(to, wethIn-amountEth);
PalmNFTXStakingZap.sol, 224, pairedToken.transfer(to, wethIn-amountEth);
NFTXFlashSwipe.sol, 51, IERC20Upgradeable(vault).transferFrom(msg.sender, address(this), mintFee + targetRedeemFee);

0xKiwi (NFTX) confirmed, but disagreed with high severity and commented:

Disagreeing with the severity, but will make sure I stick to safeTransferFrom, thank you.

0xKiwi (NFTX) resolved

LSDan (judge) decreased severity to medium and commented:

This is medium risk, not high. Loss of funds requires external factors.

Disclosures

C4 is an open organization governed by participants in the community.

C4 Contests incentivize the discovery of exploits, vulnerabilities, and bugs in smart contracts. Security researchers are rewarded at an increasing rate for finding higher-risk issues. Contest submissions are judged by a knowledgeable security researcher and solidity developer and disclosed to sponsoring developers. C4 does not conduct formal verification regarding the provided code but instead provides final verification.

C4 does not provide any guarantee or warranty regarding the security of this project. All smart contract software should be used at the sole risk and responsibility of users.