Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/264
0x486776, 0x73696d616f, 1337, 4rdiii, AgileJune, AlexCzm, ArsenLupin, BiasedMerc, Brenzee, CodeWasp, ComposableSecurity, Dots, Kalogerone, KupiaSec, Shaheen, Varun_05, ast3ros, bareli, blockchain555, cducrest-brainbot, cu5t0mPe0, durov, ff, fugazzi, funkornaut, gesha17, gladiator111, jennifer37, joicygiore, juan, kennedy1030, nisedo, sammy, techOptimizor, trachev, ubl4nk, valentin2304, xiaoming90, y4y
Users can exploit the batch minting feature to avoid paying minting fees for tokens, leading to a loss of fees for the creator and fee recipients.
File: Edition.sol 304: function mintBatch( 305: address[] calldata receivers_, 306: uint256 tokenId_, 307: uint256 amount_, 308: bytes calldata data_ 309: ) external payable { 310: // wake-disable-next-line reentrancy 311: FEE_MANAGER.collectMintFee{value: msg.value}( 312: this, tokenId_, amount_, msg.sender, address(0), works[tokenId_].strategy 313: ); 314: 315: for (uint256 i = 0; i < receivers_.length; i++) { 316: _issue(receivers_[i], tokenId_, amount_, data_); 317: } 318: 319: _refundExcess(); 320: }
Assume that the total mint fee (protocol flat fee = 0.0006 ETH + minting fee = 0.0004) for each token is 0.001 ETH. Bob wants to mint 1000 tokens, and he has to pay a minting fee of 1 ETH.
However, the issue is that Bob can mint 1000 tokens without paying the full mint fee of 1 ETH as shown in the step below:
Bob calls the above mintBatch function with receivers_ array set to 1000 x Bob address and the amount_ parameter set to 1.
Line 331 above will compute the mint fee that Bob needs to pay. In this case, the amount_ is one (1), and the total mint fee will be equal to 1 X 0.001 ETH.
The for-loop at Line 315 above will loop 1000 times as the receivers_.length is 1000. Each loop will mint one token to Bob.
At the end of the for loop, Bob will receive 1000 while only paying a minting fee of 0.001 ETH instead of 1 ETH.
In other words, Bob only paid the minting fee for the first token and did not pay the minting fee for the rest of the 999 tokens.
Loss of fees for the creator and fee recipients as minters can avoid paying the fee using the trick mentioned in this report.
Manual Review
Consider the following change to ensure that the minting fee is computed based on the total number of tokens minted to all receivers.
function mintBatch( address[] calldata receivers_, uint256 tokenId_, uint256 amount_, bytes calldata data_ ) external payable { // wake-disable-next-line reentrancy FEE_MANAGER.collectMintFee{value: msg.value}( - this, tokenId_, amount_, msg.sender, address(0), works[tokenId_].strategy + this, tokenId_, amount_ * receivers_.length, msg.sender, address(0), works[tokenId_].strategy ); for (uint256 i = 0; i < receivers_.length; i++) { _issue(receivers_[i], tokenId_, amount_, data_); } _refundExcess(); }
pqseags
Valid, will fix
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/265
14si2o_Flint, Varun_05, mt030d, xiaoming90
Original collection referrers will be overwritten when a new collection/work is created. This results in the original collection referrers being unable to collect the fee they are entitled to, leading to a loss of assets for them.
[!NOTE]
The following information is taken from the sponsor's response in the Contest's Discord Channel.
There are two types of referrers in this ecosystem. The first is a collection referrer, who refers the creation of a new Edition, and gets a cut of all mint fees for that Edition. The second is a mint referrer, who refers a mint and gets a cut of the mint fee for that mint.
The following describes the difference between collection referrer and mint referrer.
- User 1 creates a referral link to create a collection
- User 2 uses that link to publish a collection
- User 3 mints that collection. For this mint, user1 is treated as the collection referrer
- User 4 creates a referral link to mint that collection
- User 5 mints that collection. For this mint, user1 is treated as the collection referrer and user4 is treated as the mint referrer
Assume that Alice creates a referral link to create a collection. Bob uses that link to publish a new collection/work called that is based on an Edition called . The collection referrer of will Alice.
Bob will call the following TitlesCore.publish function with the edition parameter set to and the referrer_ parameter will be automatically set to Alice by the front end.
File: TitlesCore.sol 098: /// @notice Publishes a new Work in the given {Edition} using the given payload. 099: /// @param edition_ The {Edition} to publish the Work in. 100: /// @param payload_ The compressed payload for publishing the Work. See {WorkPayload}. 101: /// @param referrer_ The address of the referrer. 102: /// @return tokenId The token ID of the new Work. 103: function publish(Edition edition_, bytes calldata payload_, address referrer_) ..SNIP.. 140: // Create the fee route for the new Work 141: // wake-disable-next-line reentrancy 142: Target memory feeReceiver = feeManager.createRoute( 143: edition_, tokenId, _attributionTargets(work_.attributions), referrer_ ..SNIP..
Within the TitlesCore.publish, the following feeManager.createRoute function will be executed internally. The feeManager.createRoute function will store Alice's wallet address within the referrers[edition_] mapping. Thus, the state of the referrers mapping will be as follows:
referrers[Edition_A] = Alice
File: FeeManager.sol 125: function createRoute( 126: IEdition edition_, 127: uint256 tokenId_, 128: Target[] calldata attributions_, 129: address referrer_ 130: ) external onlyOwnerOrRoles(ADMIN_ROLE) returns (Target memory receiver) { ..SNIP.. 158: _feeReceivers[getRouteId(edition_, tokenId_)] = receiver; 159: referrers[edition_] = referrer_; 160: }
When someone mints a new token for , Alice, who is the collection referrer, will get a share of the minting fee per Line 421 below.
File: FeeManager.sol 412: function _splitProtocolFee( 413: IEdition edition_, 414: address asset_, 415: uint256 amount_, 416: address payer_, 417: address referrer_ 418: ) internal returns (uint256 referrerShare) { 419: // The creation and mint referrers earn 25% and 50% of the protocol's share respectively, if applicable 420: uint256 mintReferrerShare = getMintReferrerShare(amount_, referrer_); 421: uint256 collectionReferrerShare = getCollectionReferrerShare(amount_, referrers[edition_]); 422: referrerShare = mintReferrerShare + collectionReferrerShare;
Let's assume Charles also creates a referral link to create a collection. David uses that link to publish a new collection/work called that is based on the same Edition called . The collection referral of will be Charles.
When the FeeManager.createRoute function is executed during the publishing of the new work/collection, Charles's wallet address will be stored within the referrers[edition_] mapping. Thus, the state of the referrers mapping will be as follows:
referrers[Edition_A] = Charles
File: FeeManager.sol 125: function createRoute( 126: IEdition edition_, 127: uint256 tokenId_, 128: Target[] calldata attributions_, 129: address referrer_ 130: ) external onlyOwnerOrRoles(ADMIN_ROLE) returns (Target memory receiver) { ..SNIP.. 158: _feeReceivers[getRouteId(edition_, tokenId_)] = receiver; 159: referrers[edition_] = referrer_; 160: }
It is important to note that the referrers[Edition_A] have been updated from Alice to Charles here. At this point onwards, if someone mints tokens for , the collection referral fee will be routed to Charles instead of Alice. Alice is the referral for , yet she does not receive the referral fee, resulting in a loss of assets for Alice.
Loss of assets as shown in the above scenario. The original collection referrers are unable to collect the fee they are entitled to, leading to a loss of assets for them.
Manual Review
Consider the following changes to ensure that the collection referral fee is routed to the correct collection referrer for each collection/work.
With the following changes, Alice will continue to receive the collection referral fee for and Bob will receive the collection referral fee for even if there are multiple collections/works for a specific Edition.
function createRoute( IEdition edition_, uint256 tokenId_, Target[] calldata attributions_, address referrer_ ) external onlyOwnerOrRoles(ADMIN_ROLE) returns (Target memory receiver) { ..SNIP.. _feeReceivers[getRouteId(edition_, tokenId_)] = receiver; - referrers[edition_] = referrer_; + referrers[getRouteId(edition_, tokenId_)] = referrer_; }
function _splitProtocolFee( IEdition edition_, + uint256 tokenId, address asset_, uint256 amount_, address payer_, address referrer_ ) internal returns (uint256 referrerShare) { // The creation and mint referrers earn 25% and 50% of the protocol's share respectively, if applicable uint256 mintReferrerShare = getMintReferrerShare(amount_, referrer_); - uint256 collectionReferrerShare = getCollectionReferrerShare(amount_, referrers[edition_]); + uint256 collectionReferrerShare = getCollectionReferrerShare(amount_, referrers[getRouteId(edition_, tokenId)]); referrerShare = mintReferrerShare + collectionReferrerShare;
pqseags
Valid, but duplicate of #59
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/267
0rpse, 0x486776, 0x73696d616f, 0xboriskataa, 1337, AlexCzm, ArsenLupin, BengalCatBalu, BiasedMerc, ComposableSecurity, FastTiger, KupiaSec, Matin, Varun_05, Yu3H0, ZdravkoHr., alexzoid, ast3ros, blockchain555, brakeless, cducrest-brainbot, cryptic, cu5t0mPe0, durov, eeshenggoh, ff, gladiator111, i3arba, ironside, jennifer37, kennedy1030, merlinboii, mt030d, popeye, radin200, sammy, techOptimizor, trachev, xiaoming90, y4y
Collection referrers will not receive their share of the minting fee, leading to a loss of assets for the collection referrers.
[!NOTE]
The following information is taken from the sponsor's response in the Contest's Discord Channel.
There are two types of referrers in this ecosystem. The first is a collection referrer, who refers the creation of a new Edition, and gets a cut of all mint fees for that Edition. The second is a mint referrer, who refers a mint and gets a cut of the mint fee for that mint.
The following describes the difference between collection referrer and mint referrer.
- User 1 creates a referral link to create a collection
- User 2 uses that link to publish a collection
- User 3 mints that collection. For this mint, user1 is treated as the collection referrer
- User 4 creates a referral link to mint that collection
- User 5 mints that collection. For this mint, user1 is treated as the collection referrer and user4 is treated as the mint referrer
Assume that Alice creates a referral link to create a collection. Bob uses that link to publish a collection called . In this case, the collection referrer of will be set to Alice.
File: TitlesCore.sol 098: /// @notice Publishes a new Work in the given {Edition} using the given payload. 099: /// @param edition_ The {Edition} to publish the Work in. 100: /// @param payload_ The compressed payload for publishing the Work. See {WorkPayload}. 101: /// @param referrer_ The address of the referrer. 102: /// @return tokenId The token ID of the new Work. 103: function publish(Edition edition_, bytes calldata payload_, address referrer_) ..SNIP.. 140: // Create the fee route for the new Work 141: // wake-disable-next-line reentrancy 142: Target memory feeReceiver = feeManager.createRoute( 143: edition_, tokenId, _attributionTargets(work_.attributions), referrer_ ..SNIP..
When the TitlesCore.publish is executed to create a new collection, the feeManager.createRoute function will be executed internally. The feeManager.createRoute function will store Alice's wallet address within the referrers[edition_] mapping. Thus, the state of the referrers mapping will be as follows:
referrers[Edition_A] = Alice
File: FeeManager.sol 125: function createRoute( 126: IEdition edition_, 127: uint256 tokenId_, 128: Target[] calldata attributions_, 129: address referrer_ 130: ) external onlyOwnerOrRoles(ADMIN_ROLE) returns (Target memory receiver) { ..SNIP.. 158: _feeReceivers[getRouteId(edition_, tokenId_)] = receiver; 159: referrers[edition_] = referrer_; 160: }
When someone mints a new token for , Alice, who is the collection referral, should get a share of the minting fee. However, based on the current implementation of the FeeManager._splitProtocolFee function below, Alice will not receive her collection referral fee. Instead, the collection referral fee is routed to the mint referrer.
Line 438 below shows that the collection referral fee is routed to the mint referrer (referrer_) instead of the collection referrer (referrers[edition_]).
File: FeeManager.sol 412: function _splitProtocolFee( 413: IEdition edition_, 414: address asset_, 415: uint256 amount_, 416: address payer_, 417: address referrer_ 418: ) internal returns (uint256 referrerShare) { 419: // The creation and mint referrers earn 25% and 50% of the protocol's share respectively, if applicable 420: uint256 mintReferrerShare = getMintReferrerShare(amount_, referrer_); // @audit-info mint referrer 421: uint256 collectionReferrerShare = getCollectionReferrerShare(amount_, referrers[edition_]); // @audit-info collection referrer 422: referrerShare = mintReferrerShare + collectionReferrerShare; ..SNIP.. 430: _route( 431: Fee({asset: asset_, amount: mintReferrerShare}), 432: Target({target: referrer_, chainId: block.chainid}), 433: payer_ 434: ); 435: 436: _route( 437: Fee({asset: asset_, amount: collectionReferrerShare}), 438: Target({target: referrer_, chainId: block.chainid}), 439: payer_ 440: ); 441: }
Following are some of the negative impacts that could lead to a loss of assets:
referrer_ parameter to their own wallet address when executing the Edition.mint function, so that the collection referral fee can be routed to their own wallet, effectively avoiding paying the collection referral fee to someone else.Manual Review
Consider the following change to ensure that the collection referral fee is routed to the collection referrer instead of the mint referrer.
function _splitProtocolFee( IEdition edition_, address asset_, uint256 amount_, address payer_, address referrer_ ) internal returns (uint256 referrerShare) { // The creation and mint referrers earn 25% and 50% of the protocol's share respectively, if applicable uint256 mintReferrerShare = getMintReferrerShare(amount_, referrer_); uint256 collectionReferrerShare = getCollectionReferrerShare(amount_, referrers[edition_]); referrerShare = mintReferrerShare + collectionReferrerShare; ..SNIP.. _route( Fee({asset: asset_, amount: collectionReferrerShare}), - Target({target: referrer_, chainId: block.chainid}), + Target({target: referrers[edition_], chainId: block.chainid}), payer_ ); }
pqseags
Valid, will fix
midori-fuse
Escalate
#72, #131, #153, #158, #176, #183, and #193 are valid excluded dupes of this issue
There are also a number of other separately escalated issues (including another esc on this issue itself) that I agree are valid dupes of this. Though I can only try so hard at mentioning them :)
sherlock-admin3
Escalate
#72, #131, #153, #158, #176, #183, and #193 are valid excluded dupes of this issue
There are also a number of other separately escalated issues (including another esc on this issue itself) that I agree are valid dupes of this. Though I can only try so hard at mentioning them :)
You've created a valid escalation!
To remove the escalation from consideration: Delete your comment.
You may delete or edit your escalation comment anytime before the 48-hour escalation window closes. After that, the escalation becomes final.
sherlock-admin3
Escalate
#393 is a valid excluded dup of this issue.
You've deleted an escalation for this issue.
MageIntern
Escalate
https://github.com/sherlock-audit/2024-04-titles-judging/issues/210 should be dup
sherlock-admin3
Escalate
https://github.com/sherlock-audit/2024-04-titles-judging/issues/210 should be dup
You've created a valid escalation!
To remove the escalation from consideration: Delete your comment.
You may delete or edit your escalation comment anytime before the 48-hour escalation window closes. After that, the escalation becomes final.
WangSecurity
For issues to become duplicates, they have to be escalated, not the main report (i.e. for #210 to become a dup, you should escalate it and not the main report). Hence, planning to reject this escalation
MageIntern
For issues to become duplicates, they have to be escalated, not the main report (i.e. for #210 to become a dup, you should escalate it and not the main report). Hence, planning to reject this escalation
What should I do? Escalation is over.
midori-fuse
For issues to become duplicates, they have to be escalated, not the main report (i.e. for #210 to become a dup, you should escalate it and not the main report). Hence, planning to reject this escalation
@WangSecurity quoting escalation rule 2
https://docs.sherlock.xyz/audits/judging/escalation-period#rules-for-escalation
You can combine multiple arguments related to the same issue into one escalation. This prevents getting a double penalty when the escalation is rejected. For example, you might argue that the issue is valid and should be duplicated with #4.
Also, you can mention multiple issues in the same escalation if they need to be duplicated together or separated from the context of the issue you are escalating. For example: 'The issues #12, #145 and #5 are all duplicates of the above issue'
My escalation is then valid and should be considered
WangSecurity
Sorry for the confusion I cause above, I saw the message that the escalation was deleted and mistakenly though it's about @midori-fuse escalation and then there was a new escalation, which I thought was the only one and asked for a single issue to be a duplicate. Excuse me for the confusion, hope for your understanding. Planning to accept both escalations.
Evert0x
Result:
High
Has Duplicates
sherlock-admin4
Escalations have been resolved successfully!
Escalation status:
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
TitleGraph causes signatures generated by common EIP712 tools to be unusableSource: https://github.com/sherlock-audit/2024-04-titles-judging/issues/74
0x73696d616f, T1MOH, ZanyBonzy, ast3ros, fugazzi, mt030d
The signature in TitleGraph.acknowledgeEdge() and TitleGraph.unacknowledgeEdge() is generated based on a digest computed from edgeId and data. However, the data bytes argument is not correctly encoded according to the EIP712 specification. Consequently, a signature generated using common EIP712 tools would not pass validation in TitleGraph.checkSignature().
According to EIP712:
The dynamic values bytes and string are encoded as a keccak256 hash of their contents.
modifier checkSignature(bytes32 edgeId, bytes calldata data, bytes calldata signature) { bytes32 digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, data))); ... }
However, the checkSignature() modifier in the TitlesGraph contract reconstructs the digest by encoding the data bytes argument without first applying keccak256 hashing.
As a result, a signature generated using common EIP712 tools (e.g. using the signTypedData function from ethers.js) would not pass validation in TitleGraph.checkSignature().
// main.js const { ethers } = require("ethers"); async function main() { const pk = "0xac0974bec39a17e36ba4a6b4d238ff944bacb478cbed5efcae784d7bf4f2ff80"; const signer = new ethers.Wallet(pk); const domain = { name: "TitlesGraph", version: '1', chainId: 31337, verifyingContract: "0x7FA9385bE102ac3EAc297483Dd6233D62b3e1496" // should match the address in foundry test }; const types = { Ack: [ { name: "edgeId", type: "bytes32" }, { name: "data", type: "bytes" }, ], }; const value = { edgeId: ethers.id("test edgeId"), data: "0xabcd" }; const signature = await signer.signTypedData(domain, types, value); console.log(signature); } main();
here we run
npm install ethers node main.js
The output is 0xab4623a7bacf25ed3d6779684f195ed63a5ed1ed46c278c107390086e74b739b35f1db213c6075dedc041d68ced3d11798d49afaf3c47743d4696c49f03037b51b
// SPDX-License-Identifier: UNLICENSED pragma solidity ^0.8.13; import {Test, console} from "forge-std/Test.sol"; import {EIP712} from "lib/solady/src/utils/EIP712.sol"; contract EIP712Test is Test, EIP712 { bytes32 public constant ACK_TYPEHASH = keccak256("Ack(bytes32 edgeId,bytes data)"); // test data bytes32 testEdgeId = keccak256("test edgeId"); bytes testData = hex"abcd"; function test_sig() public { uint256 pk = 0xac0974bec39a17e36ba4a6b4d238ff944bacb478cbed5efcae784d7bf4f2ff80; bytes32 digest = _computeDigest(testEdgeId, testData); (uint8 v, bytes32 r, bytes32 s) = vm.sign(pk, digest); bytes memory signature = abi.encodePacked(r, s, v); console.logBytes(signature); } function test_sigShouldBe() public { uint256 pk = 0xac0974bec39a17e36ba4a6b4d238ff944bacb478cbed5efcae784d7bf4f2ff80; bytes32 digest = _computeDigestShouldBe(testEdgeId, testData); (uint8 v, bytes32 r, bytes32 s) = vm.sign(pk, digest); bytes memory signature = abi.encodePacked(r, s, v); console.logBytes(signature); } function _computeDigest(bytes32 edgeId, bytes memory data) internal returns (bytes32 digest) { digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, data))); } function _computeDigestShouldBe(bytes32 edgeId, bytes memory data) internal returns (bytes32 digest) { digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, keccak256(data)))); } function _domainNameAndVersion() internal pure override returns (string memory name, string memory version) { name = "TitlesGraph"; version = "1"; } }
here we run
forge test --mc EIP712Test -vv
The output is
[PASS] test_sig() (gas: 12176) Logs: 0x7bd09aece710ef3845f26c4a695d357b1b170f75d0702f18ec09409f571260237a38e0fed802f8a9d598d9aed0d7898562c51e09bfa7cf254e5a8a5bc74106561c [PASS] test_sigShouldBe() (gas: 11958) Logs: 0xab4623a7bacf25ed3d6779684f195ed63a5ed1ed46c278c107390086e74b739b35f1db213c6075dedc041d68ced3d11798d49afaf3c47743d4696c49f03037b51b
test_sig() simulates the way the digest is reconstructed in TitleGraph.checkSignature(), while test_sigShouldBe() shows how the digest should be reconstructed.
From the above output, we can see the signature generated by ethers.js matches the signature generated in test_sigShouldBe() and does not match the signature generated in test_sig().
This PoC shows the way TitleGraph.checkSignature() reconstruct the digest is not compatible with the way data is encoded in EIP712.
A signature generated by the signer using common EIP712 tools (e.g. signTypedData in ethers.js) would not pass validation in TitleGraph.checkSignature().
Manual Review, ethers.js, foundry
Encoding the data bytes as a keccak256 hash of its contents before computing the digest from it:
- digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, data))); + digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, keccak256(data))));
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/91
The protocol has acknowledged this issue.
ArsenLupin, Kalogerone, me_na0mi
zkSync Era chain has differences in the usage of the create opcode compared to the EVM.
According to the contest README, the protocol can be deployed in zkSync Era (README)
The zkSync Era docs explain how it differs from Ethereum.
The description of CREATE and CREATE2 (zkSynce Era Docs) states that Create cannot be used for arbitrary code unknown to the compiler.
TitlesCore::createEdition function uses Solady's LibClone::clone function:
function createEdition(bytes calldata payload_, address referrer_) external payable returns (Edition edition) { EditionPayload memory payload = abi.decode(payload_.cdDecompress(), (EditionPayload)); @> edition = Edition(editionImplementation.clone()); // wake-disable-next-line reentrancy edition.initialize( feeManager, graph, payload.work.creator.target, address(this), payload.metadata ); // wake-disable-next-line unchecked-return-value _publish(edition, payload.work, referrer_); emit EditionCreated( address(edition), payload.work.creator.target, payload.work.maxSupply, payload.work.strategy, abi.encode(payload.metadata) ); }
function clone(uint256 value, address implementation) internal returns (address instance) { /// @solidity memory-safe-assembly assembly { mstore(0x21, 0x5af43d3d93803e602a57fd5bf3) mstore(0x14, implementation) mstore(0x00, 0x602c3d8160093d39f33d3d3d3d363d3d37363d73) @> instance := create(value, 0x0c, 0x35) if iszero(instance) { mstore(0x00, 0x30116425) // `DeploymentFailed()`. revert(0x1c, 0x04) } mstore(0x21, 0) // Restore the overwritten part of the free memory pointer. } }
As mentioned by the zkSync docs: "The code will not function correctly because the compiler is not aware of the bytecode beforehand".
This will result in loss of funds, since there is a fee to create a new edition, hence the TitlesCore::createEdition function is payable.
No editions can be created in the zkSync Era chain.
https://github.com/sherlock-audit/2024-04-titles/blob/d7f60952df22da00b772db5d3a8272a988546089/wallflower-contract-v2/src/TitlesCore.sol#L79
https://github.com/Vectorized/solady/blob/91d5f64b39a4d20a3ce1b5e985103b8ea4dc1cfc/src/utils/LibClone.sol#L137
Manual Review
This can be solved by implementing CREATE2 directly and using type(Edition).creationCode as mentioned in the zkSync Era docs.
midori-fuse
Escalate
This escalation raises two points:
The first point is that #353 and #440 are duplicates of this issue.
The second point is that this issue has been excluded without reasoning (nor do I see any signs of someone reviewing this issue), even though evidence has been provided, and the impacts have been shown:
For this reason I believe this issue (and the mentioned duplicates) to be a valid medium.
sherlock-admin3
Escalate
This escalation raises two points:
The first point is that #353 and #440 are duplicates of this issue.
The second point is that this issue has been excluded without reasoning (nor do I see any signs of someone reviewing this issue), even though evidence has been provided, and the impacts have been shown:
- The relevant part of the zkEVM docs has been linked within the report.
- The contract is bricked for this issue, which fits into the "rendering the contract useless" in the medium severity criteria.
- The contest README includes zkSync as a deployment chain.
For this reason I believe this issue (and the mentioned duplicates) to be a valid medium.
You've created a valid escalation!
To remove the escalation from consideration: Delete your comment.
You may delete or edit your escalation comment anytime before the 48-hour escalation window closes. After that, the escalation becomes final.
cvetanovv
I disagree with the escalation. This issue is invalid.
You can see Sherlock documentation subclause 24. EVM opcodes type of issue is not valid.
midori-fuse
The clause states:
Using Solidity versions that support EVM opcodes that don't work on networks on which the protocol is deployed is not a valid issue because one can manage compilation flags to compile for past EVM versions on newer Solidity versions.
This is not an issue of "Using solidity versions", because all solidity versions support the CREATE opcode, and this issue will manifest no matter what Solidity version you use, or what EVM you compile it on. The clause clearly states that the rule only applies to using Solidity versions such that it is mitigatable by managing compilation flags, it does not apply to all "EVM opcodes type" in general.
Any issue also boils down to just "EVM opcode", because when you compile any Solidity source files, you always get EVM opcodes, the majority (if not, most of the time, all) of which you cannot manage compilation flags to the behavior you desire.
Kalogerone
I agree with @midori-fuse. Subclause 24 doesn't refer to these kind of issues. This issue is not about the solidity version used. Subclause 24 clearly covers the solution which cannot be applied here:
because one can manage compilation flags to compile for past EVM versions on newer Solidity versions.
cvetanovv
@midori-fuse and @Kalogerone, I agree with the comments and will consult with @WangSecurity about whether it is a valid issue.
Hash01011122
Agreed with @midori-fuse on
The clause clearly states that the rule only applies to using Solidity versions such that it is mitigatable by managing compilation flags, it does not apply to all "EVM opcodes type" in general.
But still this appears to be low. I will let head of judge decide severity of this issue.
Kalogerone
I would argue it is not a low because as I mentioned in my report, the function used to create a new contract is payable and the user has to pay a fee to do that. Zksync docs never mention that the function will fail/revert, meaning that the user’s funds are not safe and there is possible loss of funds.
The following code will not function correctly because the compiler is not aware of the bytecode beforehand
Unfortunately, it's impossible to differentiate between the above cases during compile-time. As a result, we strongly recommend including tests for any factory that deploys child contracts using type(T).creationCode.
WangSecurity
I agree with the escalation and that it should be a valid medium, rather than the high. Planning to accept the escalation and validate as medium with duplicates #353 and #440.
Evert0x
Result:
Medium
Has Duplicates
sherlock-admin4
Escalations have been resolved successfully!
Escalation status:
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/144
ComposableSecurity, cducrest-brainbot, sammy
In Edition.sol, functions that set the value of works[tokenId].strategy which includes works[tokenId].strategy.royaltyBps do not set ERC2981's internal token royalty value.
The function setFeeStrategy() sets the public mapping value works[tokenId_].strategy which may update the roylatyBps value but does not call _setTokenRoyalty(...):
function setFeeStrategy(uint256 tokenId_, Strategy calldata strategy_) external { if (msg.sender != works[tokenId_].creator) revert Unauthorized(); works[tokenId_].strategy = FEE_MANAGER.validateStrategy(strategy_); // @audit does not set royalties }
Similarly, the publish() function to create a new work sets the strategy but does not call _setTokenRoyalty():
function publish( address creator_, uint256 maxSupply_, uint64 opensAt_, uint64 closesAt_, Node[] calldata attributions_, Strategy calldata strategy_, Metadata calldata metadata_ ) external override onlyRoles(EDITION_MANAGER_ROLE) returns (uint256 tokenId) { tokenId = ++totalWorks; _metadata[tokenId] = metadata_; works[tokenId] = Work({ creator: creator_, totalSupply: 0, maxSupply: maxSupply_, opensAt: opensAt_, closesAt: closesAt_, strategy: FEE_MANAGER.validateStrategy(strategy_) }); Node memory _node = node(tokenId); for (uint256 i = 0; i < attributions_.length; i++) { // wake-disable-next-line reentrancy, unchecked-return-value GRAPH.createEdge(_node, attributions_[i], attributions_[i].data); } emit Published(address(this), tokenId); }
This latter case may be less of a problem since TitlesCore calls edition_.setRoyaltyTarget() right after publishing a new work. However it remains a problem if publishers are expected to interact directly with Edition and not only through TitlesCore
The value returned by works[tokenId].strategy.royaltyBps and ERC2981.royaltyInfo(tokenId, salePrice) will not be coherent. Users may expect to set a certain royalty bps while the value is not updated. Core values used for royalty payments may become incorrect after updates.
https://github.com/vectorized/solady/blob/main/src/tokens/ERC2981.sol
Manual Review
Call _setTokenRoyalty(tokenId, FeeManager.feeReceiver(address(this), tokenId), works[tokenId].strategy.royaltyBps); at the end of setFeeStrategy().
pqseags
This is valid and should be fixed
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/148
0x73696d616f, KupiaSec, fibonacci, fugazzi, juan, sammy
The FeeManager contract lacks an interface for the admin to grant or revoke roles.
The contest's README states:
ADMIN_ROLE ... On FeeManager, this role can: ... 3) Grant or revoke any role to/from any address (`grantRole`, `revokeRole`).
The FeeManager contract, inherited from solady's OwnableRoles contract, only permits the owner to manage roles.
/// @dev Allows the owner to grant `user` `roles`. /// If the `user` already has a role, then it will be a no-op for the role. function grantRoles(address user, uint256 roles) public payable virtual onlyOwner { _grantRoles(user, roles); } /// @dev Allows the owner to remove `user` `roles`. /// If the `user` does not have a role, then it will be a no-op for the role. function revokeRoles(address user, uint256 roles) public payable virtual onlyOwner { _removeRoles(user, roles); }
The FeeManager contract itself does not provide any interfaces that enable the admin to grant or revoke roles.
The admin cannot grant or revoke any roles. This limitation cannot be changed since the contract is not upgradable. A new role can only be granted or revoked by the owner, which is the TitlesCore contract, and it also lacks the corresponding functions.
https://github.com/sherlock-audit/2024-04-titles/blob/main/README.md#q-are-there-any-protocol-roles-please-list-them-and-provide-whether-they-are-trusted-or-restricted-or-provide-a-more-comprehensive-description-of-what-a-role-can-and-cant-doimpact
https://github.com/sherlock-audit/2024-04-titles/blob/main/wallflower-contract-v2/src/fees/FeeManager.sol#L115
Manual Review
Implement admin interfaces to manage roles
function grantRoles(address guy, uint256 roles) public payable override onlyOwnerOrRoles(ADMIN_ROLE) { _grantRoles(guy, roles); } function revokeRoles(address guy, uint256 roles) public payable override onlyOwnerOrRoles(ADMIN_ROLE) { _removeRoles(guy, roles); }
0xf1b0
Escalate
I believe this is a valid issue. The implementation does not align with what is described in the README.
sherlock-admin3
Escalate
I believe this is a valid issue. The implementation does not align with what is described in the README.
You've created a valid escalation!
To remove the escalation from consideration: Delete your comment.
You may delete or edit your escalation comment anytime before the 48-hour escalation window closes. After that, the escalation becomes final.
WangSecurity
Agree with escalation that it's valid, but I believe it could be spotted on deployment and fixed with re-deploy without losing funds. Moreover, the roles can still be set and revoked. Hence, believe medium is approriate here, planning to accept the escalation.
WangSecurity
Again, agree with escalation and making it a valid medium. Planning to make a new family of issues with the core issue "roles are set incorrectly or not set at all" with the following duplicates:
#166, #197, #213, #146.
0xjuaan
hi @WangSecurity, #240 is a dupe of this
Evert0x
Result:
Medium
Has Duplicates
sherlock-admin4
Escalations have been resolved successfully!
Escalation status:
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/212
4rdiii, AlexCzm, CodeWasp, KupiaSec, alexzoid, brakeless, cducrest-brainbot, den_sosnovskyi, fugazzi, i3arba, ironside, kennedy1030, radin200, recursiveEth, ubl4nk
When acknowledgeEdge() is called, the downstream call to the _setAcknowledged() method caches edges[edgeId_] in memory, instead of storage, which does not preserve changes to the Edge struct after the transaction concludes.
function acknowledgeEdge(bytes32 edgeId_, bytes calldata data_) external override returns (Edge memory edge) { if (!_isCreatorOrEntity(edges[edgeId_].to, msg.sender)) revert Unauthorized(); return _setAcknowledged(edgeId_, data_, true); } function _setAcknowledged(bytes32 edgeId_, bytes calldata data_, bool acknowledged_) internal returns (Edge memory edge) { if (!_edgeIds.contains(edgeId_)) revert NotFound(); edge = edges[edgeId_]; edge.acknowledged = acknowledged_; // @audit does this actually write to storage? the state isn't saved? if (acknowledged_) { emit EdgeAcknowledged(edge, msg.sender, data_); } else { emit EdgeUnacknowledged(edge, msg.sender, data_); } }
In the code snippet above, edges[edgeId_] is cached in edge, then modified.
The issue is that edge, the return variable, is marked as memory, which does not save the state after the transaction ends.
Therefore, the acknowledged value never changes and the acknowledgeEdge() methods do not work as intended.
Add the following test to TitlesGraph.t.sol.
Run with the following command: forge test --match-test test_acknowledgeEdgeFailure -vvvv
function test_acknowledgeEdgeFailure() public { Node memory from = Node({ nodeType: NodeType.COLLECTION_ERC1155, entity: Target({target: address(1), chainId: block.chainid}), creator: Target({target: address(2), chainId: block.chainid}), data: "" }); Node memory to = Node({ nodeType: NodeType.TOKEN_ERC1155, entity: Target({target: address(3), chainId: block.chainid}), creator: Target({target: address(4), chainId: block.chainid}), data: abi.encode(42) }); // Only the `from` node's entity can create the edge. vm.prank(from.entity.target); titlesGraph.createEdge(from, to, ""); vm.expectEmit(true, true, true, true); emit IEdgeManager.EdgeAcknowledged( Edge({from: from, to: to, acknowledged: true, data: ""}), to.creator.target, "" ); // Only the `to` node's creator (or the entity itself) can acknowledge it vm.prank(to.creator.target); titlesGraph.acknowledgeEdge(keccak256(abi.encode(from, to)), ""); (Node memory nodeTo, Node memory nodeFrom, bool ack, bytes memory dataResult ) = titlesGraph.edges(titlesGraph.getEdgeId(from, to)); console.log(ack); // edge[edgeId].acknowledged should be set to true after successful call to titlesGraph.acknowledgeEdge() // However, the value is still false. assert(true == ack); }
[FAIL. Reason: panic: assertion failed (0x01)] test_acknowledgeEdgeFailure() (gas: 429897) Logs: false
Edges are unable to be acknowledged.
Manual Review
The simplest fix would be to change memory to storage.
function _setAcknowledged(bytes32 edgeId_, bytes calldata data_, bool acknowledged_) internal - returns (Edge memory edge) + returns (Edge storage edge) { if (!_edgeIds.contains(edgeId_)) revert NotFound(); edge = edges[edgeId_]; edge.acknowledged = acknowledged_; if (acknowledged_) { emit EdgeAcknowledged(edge, msg.sender, data_); } else { emit EdgeUnacknowledged(edge, msg.sender, data_); } }
pqseags
This will be fixed
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/261
The protocol has acknowledged this issue.
0rpse, ComposableSecurity, xiaoming90
The minting process might be DOS by malicious fee recipients, breaking the protocol's core functionality. This would also result in the loss of fee for the rest of the innocent fee recipients as the minting cannot proceed, resulting in the minting fee not being collected.
When minting a new token for a given work, the minting fee will be collected as per Line 236 below.
File: Edition.sol 222: /// @notice Mint a new token for the given work. 223: /// @param to_ The address to mint the token to. 224: /// @param tokenId_ The ID of the work to mint. 225: /// @param amount_ The amount of tokens to mint. 226: /// @param referrer_ The address of the referrer. 227: /// @param data_ The data associated with the mint. Reserved for future use. 228: function mint( 229: address to_, 230: uint256 tokenId_, 231: uint256 amount_, 232: address referrer_, 233: bytes calldata data_ 234: ) external payable override { 235: // wake-disable-next-line reentrancy 236: FEE_MANAGER.collectMintFee{value: msg.value}( 237: this, tokenId_, amount_, msg.sender, referrer_, works[tokenId_].strategy 238: ); 239: 240: _issue(to_, tokenId_, amount_, data_); 241: _refundExcess(); 242: }
The collected minting fee will be routed or transferred to one or more of the following parties directly (Using a push mechanism):
This approach introduced the risk of DOS to the minting process. As long as any of the parties above intentionally revert upon receiving the ETH minting fee, they could DOS the entire minting process of work.
Note: The parties using the 0xSplit wallet are not an issue because the wallet requires the recipients to claim the fees from the 0xSplit wallet manually (Using the pull mechanism instead)
The minting process might be DOS by malicious fee recipients, breaking the protocol's core functionality. This would also result in the loss of fee for the rest of the innocent fee recipients as the minting cannot proceed, resulting in the minting fee not being collected.
Manual Review
Consider adopting a pull mechanism for the fee recipients to claim their accrued fee instead of transferring the fee directly to them during the minting process.
xiaoming9090
Escalate
The two (2) actors mentioned in this report, the "Collection referrer" and "Minting referrer," are external parties of the protocol that are not fully trusted. When these two external actors behave maliciously, they could negatively impact the other innocent parties (Edition's creator, Edition's attributions, Protocol fee receiver), including the protocol, by preventing them from receiving their fee, resulting in a loss of assets for the victims.
As such, this issue should not be overlooked due to its risks and should be considered valid.
sherlock-admin3
Escalate
The two (2) actors mentioned in this report, the "Collection referrer" and "Minting referrer," are external parties of the protocol that are not fully trusted. When these two external actors behave maliciously, they could negatively impact the other innocent parties (Edition's creator, Edition's attributions, Protocol fee receiver), including the protocol, by preventing them from receiving their fee, resulting in a loss of assets for the victims.
As such, this issue should not be overlooked due to its risks and should be considered valid.
You've created a valid escalation!
To remove the escalation from consideration: Delete your comment.
You may delete or edit your escalation comment anytime before the 48-hour escalation window closes. After that, the escalation becomes final.
WangSecurity
Agree with the escalation, even though it's pure griefing and the referrers lose fees themselves, the protocol also loses fee. Planning to accept and validate the report.
Evert0x
Result:
Medium
Has Duplicates
sherlock-admin4
Escalations have been resolved successfully!
Escalation status:
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/269
0rpse, 0x486776, 0xboriskataa, 1337, AlexCzm, ComposableSecurity, Dots, FastTiger, KupiaSec, Shaheen, Varun_05, Yu3H0, ZdravkoHr., alexzoid, ast3ros, cducrest-brainbot, cryptic, cu5t0mPe0, fugazzi, funkornaut, gladiator111, jecikpo, jennifer37, joicygiore, juan, kennedy1030, kn0t, merlinboii, mt030d, radin200, smbv-1923, techOptimizor, trachev, ubl4nk, valentin2304, xiaoming90, y4y, ydlee, zoyi
Excess ETH will be stuck in the Fee Manager contract and not swept back to the users.
Per Line 241 below, it is expected that there will be excess ETH residing on the contract at the end of the transaction. The _refundExcess function is implemented with the intention of sweeping excess ETH back to the caller of the mint function at the end of the transaction.
Assume Bob transfers 0.05 ETH to the Edition contract, but the minting fee ends up being only 0.03 ETH. The _refundExcess function at the end of the function (Line 241 below) is expected to return the excess 0.02 ETH back to Bob.
However, it was found that such a design does not work. When the collectMintFee function is executed on Line 236 below, the entire amount of ETH (0.05 ETH) will be forwarded to the Fee Manager contract. 0.03 ETH out of 0.05 ETH will be forwarded to the fee recipients, while the remaining 0.02 will be stuck in the Fee Manager contract. The excess 0.02 is not being returned to the Edition contract. Thus, when the _refundExcess function is triggered at the end of the function, no ETH will be returned to Bob.
File: Edition.sol 228: function mint( 229: address to_, 230: uint256 tokenId_, 231: uint256 amount_, 232: address referrer_, 233: bytes calldata data_ 234: ) external payable override { 235: // wake-disable-next-line reentrancy 236: FEE_MANAGER.collectMintFee{value: msg.value}( 237: this, tokenId_, amount_, msg.sender, referrer_, works[tokenId_].strategy 238: ); 239: 240: _issue(to_, tokenId_, amount_, data_); 241: _refundExcess(); 242: }
[!IMPORTANT]
This issue also affects the
Edition.mintWithCommentandEdition.mintBatchfunctions. The write-up is the same and is omitted for brevity.
In addition, the Contest's README mentioned that the protocol aims to aims to avoid any direct TVL in this release:
Please discuss any design choices you made.
Fund Management: We chose to delegate fee payouts to 0xSplits v2. The protocol aims to avoid any direct TVL in this release.
In other words, this means that no assets should be locked within the protocol. However, as shown in the earlier scenario, some assets will be stored in the Fee Manager, breaking this requirement.
Excess ETH will be stuck in the Fee Manager contract and not swept back to the users.
Manual Review
Consider forwarding only the required amount of the minting fee to the Fee Manager, so that any excess ETH can be sweeped by the _refundExcess() function at the end of the transaction.
function mint( address to_, uint256 tokenId_, uint256 tokenId_, address referrer_, bytes calldata data_ ) external payable override { + uint256 mintFee = FEE_MANAGER.getMintFee(this, tokenId_, amount_); // wake-disable-next-line reentrancy - FEE_MANAGER.collectMintFee{value: msg.value}( + FEE_MANAGER.collectMintFee{value: mintFee}( this, tokenId_, amount_, msg.sender, referrer_, works[tokenId_].strategy ); _issue(to_, tokenId_, amount_, data_); _refundExcess(); }
pqseags
Will address
WangSecurity
Firstly, the mintFee is set by the publisher/creator of the work and can be changed at any time. It's safe to assume it's highly unlikely that they will change it every (few) block(s). Moreover, Edition contract has to functions to get the mintFee here and here. Hence, the user may know exactly how much msg.value the need to send.
However, it doesn't exclude the situation when the mintFee is change in the exact same block right before the user mints, resulting in the revert. But the user can mitigate it on their side: call mintFee and mint in the same call.
Therefore, I believe medium severity is more appropriate here, planning to downgrade the issue.
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/273
The protocol has acknowledged this issue.
0x73696d616f, 1337, Bigsam, ComposableSecurity, Krace, KupiaSec, Yu3H0, alexzoid, cryptic, fibonacci, fugazzi, jecikpo, jennifer37, juan, kennedy1030, mt030d, xiaoming90, zoyi
Malicious users can block someone from acknowledging or deacknowledging an edge, affecting the sanctity of the data in the Graph.
[!IMPORTANT]
The following is an extract from the contest's README:
Additional audit information.
In addition to the security of funds, we would also like there to be focus on the sanctity of the data in the TitlesGraph and the permissioning around it (only the appropriate people/contracts can signal reference and acknowledgement of reference).
The contest's README stated that apart from the loss of assets, the protocol team would like there to be a focus on the sanctity of the data. Thus, any issues related to the sanctity of the data in the Graph would be considered a valid Medium issue in the context of this audit contest, as the Contest's README supersede Sherlock's judging rules per Sherlock's Hierarchy of truth.
Both acknowledgeEdge and unacknowledgeEdge functions rely on the same modifier (checkSignature) to verify the signature validity. Thus, the signature used for acknowledgment and deacknowledgment of an edge follows the same format and can be used interchangeably. However, this design creates an issue, as described next.
Assume that Bob wants to acknowledge an edge. Thus, Bob calls the acknowledgeEdge function with his signature .
A malicious user can always front-run Bob, take his signature () and sent to the unacknowledgeEdge function instead. When the unacknowledgeEdge function is executed with , the signature will be marked as used by the code _isUsed[keccak256(signature)] = true;.
When Bob's acknowledgeEdge transaction gets executed, it will revert because his signature () has already been used.
The malicious users can keep repeating as the chain's gas fees on L2 are cheap.
The same trick can also be used to block someone from deacknowledge an edge.
File: TitlesGraph.sol 118: function acknowledgeEdge(bytes32 edgeId_, bytes calldata data_, bytes calldata signature_) 119: external 120: checkSignature(edgeId_, data_, signature_) 121: returns (Edge memory edge) 122: { 123: return _setAcknowledged(edgeId_, data_, true); 124: }
File: TitlesGraph.sol 146: function unacknowledgeEdge(bytes32 edgeId_, bytes calldata data_, bytes calldata signature_) 147: external 148: checkSignature(edgeId_, data_, signature_) 149: returns (Edge memory edge) 150: { 151: return _setAcknowledged(edgeId_, data_, false); 152: }
File: TitlesGraph.sol 39: /// @notice Modified to check the signature for a proxied acknowledgment. 40: modifier checkSignature(bytes32 edgeId, bytes calldata data, bytes calldata signature) { 41: bytes32 digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, data))); 42: if ( 43: !edges[edgeId].to.creator.target.isValidSignatureNowCalldata(digest, signature) 44: || _isUsed[keccak256(signature)] 45: ) { 46: revert Unauthorized(); 47: } 48: _; 49: _isUsed[keccak256(signature)] = true; 50: }
Malicious users can block someone from acknowledging or deacknowledging an edge, affecting the sanctity of the data in the Graph.
Manual Review
bytes32 public constant ACK_TYPEHASH = keccak256("Ack(bytes32 edgeId,bytes data)"); bytes32 public constant DEACK_TYPEHASH = keccak256("Deack(bytes32 edgeId,bytes data)"); bytes32 digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, data))); bytes32 digest = _hashTypedData(keccak256(abi.encode(DEACK_TYPEHASH, edgeId, data)));
Consider using two different hash types for acknowledging or deacknowledging within the signature and use a different modifier for checking the signature. This will prevent malicious users from taking the signature intended for acknowledgeEdge and submitting it to unacknowledgeEdge, and vice versa.
modifier checkSignatureForAck(bytes32 edgeId, bytes calldata data, bytes calldata signature) { bytes32 digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, data))); ..SNIP.. } modifier checkSignatureForDeack(bytes32 edgeId, bytes calldata data, bytes calldata signature) { bytes32 digest = _hashTypedData(keccak256(abi.encode(DEACK_TYPEHASH, edgeId, data))); ..SNIP.. }
function acknowledgeEdge(bytes32 edgeId_, bytes calldata data_, bytes calldata signature_) external - checkSignature(edgeId_, data_, signature_) + checkSignatureForAck(edgeId_, data_, signature_) returns (Edge memory edge) { function unacknowledgeEdge(bytes32 edgeId_, bytes calldata data_, bytes calldata signature_) external - checkSignature(edgeId_, data_, signature_) + checkSignatureForDeack(edgeId_, data_, signature_) returns (Edge memory edge) {
With this design, if a creator creates a signature intended for acknowledgeEdge function, and a malicious user front-runs and copies the signature and submits it to acknowledgeEdge function, no harm is done as the malicious user is simply executing the task on behalf of the creator. The edge will still be acknowledged at the end.
pqseags
Will investigate
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/279
The protocol has acknowledged this issue.
0x73696d616f, 1337, ComposableSecurity, T1MOH, Yu3H0, ZdravkoHr., cducrest-brainbot, fibonacci, juan, xiaoming90, zigtur
The signature is malleable. When a signature is malleable, it means that it is possible to produce another valid signature for the same message (which also means the same digest).
If the intention is only to allow the creator to acknowledge an edge once, the creator can bypass this restriction because the signature is malleable, and the creator can create another signature to acknowledge an edge again.
The protocol relies on Solady's SignatureCheckerLib to verify that the signature provided is valid. Once a signature has been successfully verified, it will be marked as used in Line 49 below to prevent users from re-using or replaying the same signature.
The Solady's SignatureCheckerLib warns that the library does not check if a signature is non-malleable.
/// WARNING! Do NOT use signatures as unique identifiers: /// - Use a nonce in the digest to prevent replay attacks on the same contract. /// - Use EIP-712 for the digest to prevent replay attacks across different chains and contracts. /// EIP-712 also enables readable signing of typed data for better user safety. /// This implementation does NOT check if a signature is non-malleable. library SignatureCheckerLib {
Based on the following code, a creator can only acknowledge an edge once because the digest of the signature to acknowledge an edge will be the same (assuming that data is not in use) every time. After a creator acknowledges an edge, the signature (which also means its digest) will be marked as used in Line 49 below, thus preventing the creator from acknowledging the edge again later using the same signature.
File: TitlesGraph.sol 39: /// @notice Modified to check the signature for a proxied acknowledgment. 40: modifier checkSignature(bytes32 edgeId, bytes calldata data, bytes calldata signature) { 41: bytes32 digest = _hashTypedData(keccak256(abi.encode(ACK_TYPEHASH, edgeId, data))); 42: if ( 43: !edges[edgeId].to.creator.target.isValidSignatureNowCalldata(digest, signature) 44: || _isUsed[keccak256(signature)] 45: ) { 46: revert Unauthorized(); 47: } 48: _; 49: _isUsed[keccak256(signature)] = true; 50: }
When a signature is malleable, it means that it is possible to produce another valid signature for the same message (which also means the same digest).
If the intention is only to allow the creator to acknowledge an edge once, the creator can bypass this restriction because the signature is malleable, and the creator can create another signature to acknowledge an edge again.
Manual Review
Consider verifying the s of the signature is within valid bounds to avoid signature malleability.
if (uint256(s) > 0x7FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF5D576E7357A4501DDFE92F46681B20A0) { revert("ECDSA: invalid signature 's' value"); }
pqseags
This seems to be a duplicate of #273 @Hash01011122
WangSecurity
Planning to make a new issue family with this report as the main one. Planning to add the following duplicated:
0xjuaan
@WangSecurity add https://github.com/sherlock-audit/2024-04-titles-judging/issues/130, #10 to that list, they are currently incorrectly dupes of #273 which is a dos/frontrunning issue
Evert0x
@WangSecurity add #130, #10 to that list, they are currently incorrectly dupes of #273 which is a dos/frontrunning issue
Thank you!
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/280
0x486776, 0xShiki, 0xlucky, AgileJune, ArsenLupin, BiasedMerc, ComposableSecurity, Dots, FastTiger, IllIllI, Kalogerone, KupiaSec, Shaheen, Varun_05, ZdravkoHr., alexzoid, ast3ros, azanux, cducrest-brainbot, cu5t0mPe0, ff, fugazzi, funkornaut, jennifer37, joicygiore, juan, kennedy1030, kn0t, mt030d, sammy, smbv-1923, trachev, xiaoming90, y4y, ydlee, zigtur
The core minting feature of the protocol is broken due to the mishandling of msg.value within the for-loop.
Assume that the total fee for each token is 0.001 ETH, and Bob wants to mint four tokens. The total fee will be 0.004 ETH, so he will send 0.004 ETH when calling the above mintBatch function.
An important point to note is that the msg.value will always remain at 0.004 ETH throughout the entire execution of the mintBatch function. The msg.value will not automatically be reduced regardless of how many ETH has been transferred out or "spent".
In the first for-loop, the msg.value will be 0.004 ETH, and all 0.004 ETH will be routed to the fee manager and subsequently routed to the fee recipient address/0xSplit wallet.
In the second for-loop, since all the ETH (0.004 ETH) was sent to the fee manager earlier, the amount of ETH left on the Edition contract is zero. When the second for-loop attempts to send msg.value (0.004 ETH) to the fee manager again, it will revert due to insufficient ETH, and the transaction will fail and revert. Thus, this batch minting feature is broken.
File: Edition.sol 277: function mintBatch( 278: address to_, 279: uint256[] calldata tokenIds_, 280: uint256[] calldata amounts_, 281: bytes calldata data_ 282: ) external payable { 283: for (uint256 i = 0; i < tokenIds_.length; i++) { 284: Work storage work = works[tokenIds_[i]]; 285: 286: // wake-disable-next-line reentrancy 287: FEE_MANAGER.collectMintFee{value: msg.value}( 288: this, tokenIds_[i], amounts_[i], msg.sender, address(0), work.strategy 289: ); 290: 291: _checkTime(work.opensAt, work.closesAt); 292: _updateSupply(work, amounts_[i]); 293: } 294: 295: _batchMint(to_, tokenIds_, amounts_, data_); 296: _refundExcess(); 297: }
Breaks core contract functionality. The batch minting feature, a core feature of the protocol, is broken.
Manual Review
For each loop, consider only forwarding/transferring the minting fee for the current token ID instead of the entire ETH (msg.value).
function mintBatch( address to_, uint256[] calldata tokenIds_, uint256[] calldata amounts_, bytes calldata data_ ) external payable { for (uint256 i = 0; i < tokenIds_.length; i++) { Work storage work = works[tokenIds_[i]]; + uint256 mintFee = FEE_MANAGER.getMintFee(work.strategy, amounts_[i]) // wake-disable-next-line reentrancy + FEE_MANAGER.collectMintFee{value: mintFee}( - FEE_MANAGER.collectMintFee{value: msg.value}( this, tokenIds_[i], amounts_[i], msg.sender, address(0), work.strategy );
pqseags
I don't think the duplicates on this issue are actually duplicates of this issue. @Hash01011122
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/283
0x73696d616f, BengalCatBalu, BiasedMerc, CodeWasp, KupiaSec, Varun_05, alexzoid, cu5t0mPe0, jennifer37, mt030d, sammy, xiaoming90, y4y, zoyi
The new creators are unable to update the royalty target and the fee route for their works. As a result, it could lead to a loss of assets for the new creator due to the following:
The creator can call the transferWork to transfer the work to another creator. However, it was observed that after the transfer, there is still much information about the work pointing to the previous creator instead of the new creator.
File: Edition.sol 412: function transferWork(address to_, uint256 tokenId_) external { 413: Work storage work = works[tokenId_]; 414: if (msg.sender != work.creator) revert Unauthorized(); 415: 416: // Transfer the work to the new creator 417: work.creator = to_; 418: 419: emit WorkTransferred(address(this), tokenId_, to_); 420: }
Following are some instances of the issues:
To aggravate the issues, creators cannot call the Edition.setRoyaltyTarget and FeeManager.createRoute as these functions can only executed by EDITION_MANAGER_ROLE and [Owner, Admin], respectively.
Specifically for the Edition.setRoyaltyTarget function that can only be executed by EDITION_MANAGER_ROLE, which is restricted and not fully trusted in the context of this audit. The new creator could have purchased the work from the previous creator, but only to find out that the malicious edition manager decided not to update the royalty target to point to the new creator's address for certain reasons, leading to the royalty fee continuing to be routed to the previous creator. In this case, it negatively impacts the new creator as it leads to a loss of royalty fee for the new creator.
[!NOTE]
The following is an extract from the contest's README stating that the EDITION_MANAGER_ROLE is restricted. This means that any issue related to EDITION_MANAGER_ROLE that could affect TITLES protocol/users negatively will be considered valid in this audit contest.
EDITION_MANAGER_ROLE (Restricted) =>
Loss of assets for the new creator due to the following:
Manual Review
Consider allowing the creators themselves to have the right to update the royalty target and the fee route for their works.
pqseags
While this is unintuitive, we do not intend to update the fee splits when ownership of a work is changed after publishing
cducrest
Escalate
Should be low/info as this is just un-intuitive behaviour of the protocol and does not represent a real threat / loss of funds.
The team does not intend to update the behaviour confirming that this is intended.
The problem arises when users transfer a work to another user while also believing it will transfer the ownership of the fees. This is a user mistake / misuse of the protocol and not an attack / exploit of a vulnerability.
sherlock-admin3
Escalate
Should be low/info as this is just un-intuitive behaviour of the protocol and does not represent a real threat / loss of funds.
The team does not intend to update the behaviour confirming that this is intended.
The problem arises when users transfer a work to another user while also believing it will transfer the ownership of the fees. This is a user mistake / misuse of the protocol and not an attack / exploit of a vulnerability.
You've created a valid escalation!
To remove the escalation from consideration: Delete your comment.
You may delete or edit your escalation comment anytime before the 48-hour escalation window closes. After that, the escalation becomes final.
Hash01011122
@pqseags @cducrest can you please elaborate on how this issue just un-intuitive behavior of the protocol? Also where was it mentioned at the time of contest that protocol does not intend to update the fee splits when ownership of a work is changed after publishing
pqseags
It was not documented, but there isn't any logic in the code that makes an attempt to modify the fee route. In our minds, changing the creator on a work is more for the sake of changing who has admin responsibility. As this wasn't documented clearly in the read-me, I'll leave it to the judges to determine severity level given the context.
cducrest
It is not possible to change the fee split whether before or after ownership of a work is changed. To me, this is an optional feature request. It is not like it was possible to change the fee / royalty target before the work was transferred to a new creator and it becomes no longer possible. It is never possible to update it.
Also it makes little sense to update the "creator" of a work. If a "creator" (in the English literal understanding) publishes a work (a book, a picture, an NFT, anything) they remain the "creator" forever. Shakespeare cannot transfer authorship of Hamlet. As mentioned by @pqseags the transferWork() function serves more administrative purpose and delegating admin rights. It does not necessarily need to transfer fee/royalty rights.
WangSecurity
Thank you for these responses and insightful examples, but I agree that this design is actuall counter-intuitive and it wasn't documented anywhere. Also, I believe examples of how it works somewhere else are irrelevent. The watsons couldn't know what was the correct design for this function or it's known that when you transfer ownership the royalty is sent to the previous owner, hence, I believe it's fair to keep it valid.
Planning to reject the escalation and leave the issue as it is.
Evert0x
Result:
Medium
Has Duplicates
sherlock-admin4
Escalations have been resolved successfully!
Escalation status:
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/285
The protocol has acknowledged this issue.
xiaoming90
Malicious EDITION_MANAGER_ROLE can front-run victims to increase royalty, leading to a loss of assets for the victim.
The following is an extract from the contest's README stating that the EDITION_MANAGER_ROLE is restricted. This means that any issue related to EDITION_MANAGER_ROLE that could affect TITLES protocol/users negatively will be considered valid in this audit contest.
EDITION_MANAGER_ROLE (Restricted) =>
On an Edition, this role can:Set the Edition's ERC2981 royalty receiver (
setRoyaltyTarget). This is the only way to change the royalty receiver for the Edition.
[!NOTE]
About ERC2981
ERC2981 known as "NFT Royalty Standard." It introduces a standardized way to handle royalty payments for non-fungible tokens (NFTs) on the Ethereum blockchain, providing a mechanism to ensure creators receive a share of the proceeds when their NFTs are resold
Assume that the current royalty for work/collection X () is 5%. When secondary marketplaces resell the TITLES NFT, they will retrieve the NFT's royalty information by calling the Edition.royaltyInfo function. The royalty payments are calculated, collected, and transferred automatically to the creator's wallet whenever their NFT is resold.
The issue is that the EDITION_MANAGER_ROLE is restricted in the context of this audit and is considered not fully trusted.
Assume Bob submits a transaction to purchase an NFT from to the mempool. A malicious editor manager could front-run the purchase transaction and attempt to increase the royalty to be increased from 5% to 95%, resulting in more royalty fees being collected from Bob, and routed to the creators. This leads to a loss of assets for Bob.
File: Edition.sol 386: /// @notice Set the ERC2981 royalty target for the given work. 387: /// @param tokenId The ID of the work. 388: /// @param target The address to receive royalties. 389: function setRoyaltyTarget(uint256 tokenId, address target) 390: external 391: onlyRoles(EDITION_MANAGER_ROLE) 392: { 393: _setTokenRoyalty(tokenId, target, works[tokenId].strategy.royaltyBps); 394: }
Loss of assets for the victim, as shown in the above scenario
Manual Review
Ensure that only trusted users can update the royalty information, as this is a sensitive value that could affect the fee being charged when TITLES NFT is resold in the secondary market.
xiaoming9090
Escalate
EDITION_MANAGER_ROLE (Restricted) =>
Any issue related to EDITION_MANAGER_ROLE that could affect TITLES protocol/users negatively will be considered valid in this audit contest. This report demonstrates the negative impacts that a malicious editor manager could cause to the users. Thus, this issue should be valid.
Per Sherlock Judging rules, note that Contest README supersedes anything else, including protocol answers on the contest public Discord channel. This Sherlock rule must be adhered to strictly.
Hierarchy of truth: Contest README > Sherlock rules for valid issues > protocol documentation (including code comments) > protocol answers on the contest public Discord channel.
sherlock-admin3
Escalate
- Per the contest's README, it states that the EDITION_MANAGER_ROLE is restricted.
EDITION_MANAGER_ROLE (Restricted) =>
Any issue related to EDITION_MANAGER_ROLE that could affect TITLES protocol/users negatively will be considered valid in this audit contest. This report demonstrates the negative impacts that a malicious editor manager could cause to the users. Thus, this issue should be valid.
Per Sherlock Judging rules, note that Contest README supersedes anything else, including protocol answers on the contest public Discord channel. This Sherlock rule must be adhered to strictly.
Hierarchy of truth: Contest README > Sherlock rules for valid issues > protocol documentation (including code comments) > protocol answers on the contest public Discord channel.
- The following issues are not duplicated of this issue as they did not demonstrate how a malicious editor manager can front-run victims to increase royalty, causing a negative impact on the users.
You've created a valid escalation!
To remove the escalation from consideration: Delete your comment.
You may delete or edit your escalation comment anytime before the 48-hour escalation window closes. After that, the escalation becomes final.
WangSecurity
Agree with the escalation. Planning to accept it and validate issue to unique medium.
Evert0x
Result:
Medium
Unique
sherlock-admin4
Escalations have been resolved successfully!
Escalation status:
ccashwell
Why is this accepted? It's a generic informational issue at best, the finding literally applies to any modifiable configuration that affects pricing in any contract. This is not a valid logic issue.
WangSecurity
Based on the README, we have to consider that Edition_Manager_Role to be restricted, hence, untrusted. In that sense, there is a possibility it will harm users as shown in this report, thus, exploiting the current logic of the protocol.
ccashwell
I think the word "exploiting" is wrong here, what's reported is that an NFT's creator can change pricing. Literally anytime this occurs is valid. Moreover, it would assume that the royalty payer (i.e. secondary market buyer) somehow sent an amount of assets beyond the expected amount due.
There are many many ways to modify configs such that users end up paying more, and "front running" is irrelevant to that. For example, if the team changed core protocol fees, could that not also qualify as malicious under such a broad interpretation? Fee changes are a normal part of the protocol (and editions), and all users are aware of the exact quoted amount at the time they click "mint". If they send more than this amount, that's user error.
Edition.supportsInterface is not EIP1155 compliantSource: https://github.com/sherlock-audit/2024-04-titles-judging/issues/287
CodeWasp, ZdravkoHr., cducrest-brainbot, xiaoming90
According to the ERC-1155 specification, the smart contracts that are implementing it MUST have a supportsInferface(bytes4) function that returns true for values 0xd9b67a26 and 0x0e89341c. The current implementation of Edition.sol will return false for both these values.
The contract inherits from ERC1155 and ERC2981.
contract Edition is IEdition, ERC1155, ERC2981, Initializable, OwnableRoles
The supportsInterface() function of Edition returns the result of executing super.supportsInterface()
function supportsInterface(bytes4 interfaceId) public view override(IEdition, ERC1155, ERC2981) returns (bool) { return super.supportsInterface(interfaceId); }
Since both ERC1155 and ERC2981 implement that function and ERC2981 is the more derived contract of the two, Edition.supportsInterface() will end up executing only ERC2981.supportsInterface().
Medium. The contract is to be a strict implementation of ERC1155, but it does not implement the mandatory ERC1155.supportsInterface() function.
PoC for Edition.t.sol
function test_interface() public { assertFalse(edition.supportsInterface(bytes4(0xd9b67a26))); assertFalse(edition.supportsInterface(bytes4(0x0e89341c))); }
Foundry
Instead of relying on super, return the union of ERC1155.supportsInterface(interfaceId) and ERC2981.supportsInterface(interfaceId).
function supportsInterface(bytes4 interfaceId) public view override(IEdition, ERC1155, ERC2981) returns (bool) { - return super.supportsInterface(interfaceId); + return ERC1155.supportsInterface(interfaceId) || ERC2981.supportsInterface(interfaceId); }
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.
Source: https://github.com/sherlock-audit/2024-04-titles-judging/issues/445
0x73696d616f, 1337, Yu3H0, alexzoid, blackhole, fibonacci, fugazzi, trachev, xiaoming90
The TitlesGraph contract is designed to be upgradeable, utilizing the UUPSUpgradeable pattern. However, it's instantiated via a constructor in the TitlesCore contract setup.
In the TitlesCore contract, TitlesGraph is instantiated directly using a constructor rather than being set up as a proxy. This could lead to unexpected behavior when attempting to upgrade the contract, as the proxy would not have access to the initialized state variables or might interact incorrectly with uninitialized storage.
Inability to leverage the upgradeability.
graph = new TitlesGraph(address(this), msg.sender);
Manual Review
Deploy the TitlesGraph contract without initializing state in the constructor. Deploy a proxy that points to the deployed TitlesGraph implementation. Correct approach using a proxy pattern for upgradeable contracts:
address graphImplementation = address(new TitlesGraph()); graph = TitlesGraph(payable(LibClone.deployERC1967(graphImplementation))); graph.initialize(address(this), msg.sender);
alexzoid-eth
Escalate
This issue is not a duplicate of #272. This is a valid medium issue uncovering the inability of TitlesGraph contract upgrade.
Possible duplicates are #87 #142 #170 #180 #209 #281 #319 #342
sherlock-admin3
Escalate
This issue is not a duplicate of #272. This is a valid medium issue uncovering the inability of
TitlesGraphcontract upgrade.Possible duplicates are #87 #142 #170 #180 #209 #281 #319 #342
You've created a valid escalation!
To remove the escalation from consideration: Delete your comment.
You may delete or edit your escalation comment anytime before the 48-hour escalation window closes. After that, the escalation becomes final.
realfugazzi
Agree with escalation. Also documented the issue in #170 which has been incorrectly duped as #272 too
Hash01011122
Responded in #272.
Borderline low/medium. Tending towards low because in earlier contest issues like this were considered low.
realfugazzi
Responded in #272.
Borderline low/medium. Tending towards low because in earlier contest issues like this were considered low.
I think you are confusing this issue with #281, which is an entirely different problem. There are at least 3 different issues being grouped here, see https://github.com/sherlock-audit/2024-04-titles-judging/issues/272#issuecomment-2113628980
alexzoid-eth
Responded in #272.
Borderline low/medium. Tending towards low because in earlier contest issues like this were considered low.
Medium due to rules (https://docs.sherlock.xyz/audits/judging/judging#v.-how-to-identify-a-medium-issue): Breaks core contract functionality.
WangSecurity
Agree with the escalation, planning to accept it and make a new issue family of medium severity with the following duplicates:
#142, #87, #170, #180, #209, #281, #319, #342, #182
Evert0x
Result:
Medium
Has Duplicates
sherlock-admin2
Escalations have been resolved successfully!
Escalation status:
sherlock-admin2
The protocol team fixed this issue in the following PRs/commits:
https://github.com/titlesnyc/wallflower-contract-v2/pull/1
sherlock-admin2
The Lead Senior Watson signed off on the fix.