Introducing Code4rena Blue: Dedicated defense. Competitive bounties. Independent judging.Learn more →

InsureDAO contest
Findings & Analysis Report

2022-03-15

Table of contents

Overview

About C4

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 InsureDAO smart contract system written in Solidity. The code contest took place between January 7—13, 2022.

Wardens

40 Wardens contributed reports to the InsureDAO contest:

  1. WatchPug (jtp and ming)
  2. leastwood
  3. cmichel
  4. danb
  5. sirhashalot
  6. Dravee
  7. p4st13r4 (0xb4bb4 and 0x69e8)
  8. pauliax
  9. camden
  10. robee
  11. hyh
  12. defsec
  13. Jujic
  14. 0x1f8b
  15. gzeon
  16. 0xngndev
  17. loop
  18. ye0lde
  19. Ruhum
  20. TomFrenchBlockchain
  21. Fitraldys
  22. cccz
  23. hubble (ksk2345 and shri4net)
  24. Tomio
  25. tqts
  26. pmerkleplant
  27. ospwner
  28. egjlmn1
  29. Meta0xNull
  30. OriDabush
  31. 0x0x0x
  32. saian
  33. solgryn
  34. csanuragjain
  35. Kumpirmafyas
  36. jayjonah8
  37. bugwriter001

This contest was judged by 0xean (judge).

Final report assembled by CloudEllie and itsmetechjay.

Summary

The C4 analysis yielded an aggregated total of 65 unique vulnerabilities and 178 total findings. All of the issues presented here are linked back to their original finding.

Of these vulnerabilities, 14 received a risk rating in the category of HIGH severity, 9 received a risk rating in the category of MEDIUM severity, and 42 received a risk rating in the category of LOW severity.

C4 analysis also identified 15 non-critical recommendations and 98 gas optimizations.

Scope

The code under review can be found within the C4 InsureDAO contest repository, and is composed of 10 smart contracts written in the Solidity programming language and includes 2445 source 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 (14)

[H-01] Tokens can be burned with no access control

Submitted by sirhashalot

The Vault.sol contract has two address state variables, the keeper variable and the controller variable, which are both permitted to be the zero address. If both variables are zero simultaneously, any address can burn the available funds (available funds = balance - totalDebt) by sending these tokens to the zero address with the unprotected utilitize() function. If a user has no totalDebt, the user can lose their entire underlying token balance because of this.

Proof of Concept

The problematic utilize() function is found here. To see how the two preconditions can occur:

  1. The keeper state variable is only changed by the setKeeper() function found here. If this function is not called, the keeper variable will retain the default value of address(0), which bypasses the only access control for the utilize function.
  2. There is a comment here on line 69 stating the controller state variable can be zero. There is no zero address check for the controller state variable in the Vault constructor.

If both address variables are left at their defaults of address(0), then the safeTransfer() call on line 348 would send the tokens to address(0).

Add the following line to the very beginning of the utilize() function: require(address(controller) != address(0))

This check is already found in many other functions in Vault.sol, including the _unutilize() function.

oishun1112 (Insure) confirmed and resolved:

https://github.com/InsureDAO/pool-contracts/blob/audit/code4rena/contracts/Vault.sol#L382

[H-02] Typo in PoolTemplate unlock function results in user being able to unlock multiple times

Submitted by loop, also found by p4st13r4 and ye0lde

The function unlock() in PoolTemplate has a typo where it compares insurances[_id].status to false rather than setting it to false. If the conditions are met to unlock the funds for an id, the user should be able to call the unlock() function once for that id as insurances[_id].amount is subtracted from lockedAmount. However, since insurances[_id].status does not get set to false, a user can call unlock() multiple times for the same id, resulting in lockedAmount being way smaller than it should be since insurances[_id].amount is subtracted multiple times.

Impact

lockedAmount is used to calculate the amount of underlying tokens available for withdrawals. If lockedAmount is lower than it should be users are able to withdraw more underlying tokens than available for withdrawals.

Proof of Concept

Typo in unlock():

Calculation of underlying tokens available for withdrawal:

Change insurances[_id].status == false; to insurances[_id].status = false;

oishun1112 (Insure) confirmed and resolved:

https://github.com/InsureDAO/pool-contracts/blob/audit/code4rena/contracts/PoolTemplate.sol#L375

0xean (judge) commented:

upgrading to sev-3 based on assets being compromised.

[H-03] Malicious Market Creators Can Steal Tokens From Unsuspecting Approved Reference Accounts

Submitted by leastwood

The current method of market creation involves calling Factory.createMarket() with a list of approved _conditions and _references accounts. If a registered template address has templates[address(_template)].isOpen == true, then any user is able to call createMarket() using this template. If the template points to PoolTemplate.sol, then a malicious market creator can abuse PoolTemplate.initialize() as it makes a vault deposit from an account that they control. The vulnerable internal function, _depositFrom(), makes a vault deposit from the _references[4] address (arbitrarily set to an approved reference address upon market creation).

Hence, if approved _references accounts have set an unlimited approval amount for Vault.sol before deploying their market, a malicious user can frontrun market creation and cause these tokens to be transferred to the incorrect market.

This issue can cause honest market creators to have their tokens transferred to an incorrectly configured market, leading to unrecoverable funds. If their approval to Vault.sol was set to the unlimited amount, malicious users will also be able to force honest market creators to transfer more tokens than they would normally want to allow.

Proof of Concept

https://github.com/code-423n4/2022-01-insure/blob/main/contracts/Factory.sol#L158-L231

function createMarket(
    IUniversalMarket _template,
    string memory _metaData,
    uint256[] memory _conditions,
    address[] memory _references
) public override returns (address) {
    //check eligibility
    require(
        templates[address(_template)].approval == true,
        "ERROR: UNAUTHORIZED_TEMPLATE"
    );
    if (templates[address(_template)].isOpen == false) {
        require(
            ownership.owner() == msg.sender,
            "ERROR: UNAUTHORIZED_SENDER"
        );
    }
    if (_references.length > 0) {
        for (uint256 i = 0; i < _references.length; i++) {
            require(
                reflist[address(_template)][i][_references[i]] == true ||
                    reflist[address(_template)][i][address(0)] == true,
                "ERROR: UNAUTHORIZED_REFERENCE"
            );
        }
    }

    if (_conditions.length > 0) {
        for (uint256 i = 0; i < _conditions.length; i++) {
            if (conditionlist[address(_template)][i] > 0) {
                _conditions[i] = conditionlist[address(_template)][i];
            }
        }
    }

    if (
        IRegistry(registry).confirmExistence(
            address(_template),
            _references[0]
        ) == false
    ) {
        IRegistry(registry).setExistence(
            address(_template),
            _references[0]
        );
    } else {
        if (templates[address(_template)].allowDuplicate == false) {
            revert("ERROR: DUPLICATE_MARKET");
        }
    }

    //create market
    IUniversalMarket market = IUniversalMarket(
        _createClone(address(_template))
    );

    IRegistry(registry).supportMarket(address(market));
    
    markets.push(address(market));


    //initialize
    market.initialize(_metaData, _conditions, _references);

    emit MarketCreated(
        address(market),
        address(_template),
        _metaData,
        _conditions,
        _references
    );

    return address(market);
}

https://github.com/code-423n4/2022-01-insure/blob/main/contracts/PoolTemplate.sol#L178-L221

function initialize(
    string calldata _metaData,
    uint256[] calldata _conditions,
    address[] calldata _references
) external override {
    require(
        initialized == false &&
            bytes(_metaData).length > 0 &&
            _references[0] != address(0) &&
            _references[1] != address(0) &&
            _references[2] != address(0) &&
            _references[3] != address(0) &&
            _references[4] != address(0) &&
            _conditions[0] <= _conditions[1],
        "ERROR: INITIALIZATION_BAD_CONDITIONS"
    );
    initialized = true;

    string memory _name = string(
        abi.encodePacked(
            "InsureDAO-",
            IERC20Metadata(_references[1]).name(),
            "-PoolInsurance"
        )
    );
    string memory _symbol = string(
        abi.encodePacked("i-", IERC20Metadata(_references[1]).symbol())
    );
    uint8 _decimals = IERC20Metadata(_references[0]).decimals();

    initializeToken(_name, _symbol, _decimals);

    registry = IRegistry(_references[2]);
    parameters = IParameters(_references[3]);
    vault = IVault(parameters.getVault(_references[1]));

    metadata = _metaData;

    marketStatus = MarketStatus.Trading;

    if (_conditions[1] > 0) {
        _depositFrom(_conditions[1], _references[4]);
    }
}

Tools Used

Manual code review. Discussions with kohshiba.

After discussions with the sponsor, they have opted to parse a _creator address to PoolTemplate.sol which will act as the depositor and be set to msg.sender in Factory.createMarket(). This will prevent malicious market creators from forcing vault deposits from unsuspecting users who are approved in Factory.sol and have also approved Vault.sol to make transfers on their behalf.

oishun1112 (Insure) confirmed:

https://github.com/code-423n4/2022-01-insure-findings/issues/250

[H-04] Initial pool deposit can be stolen

Submitted by cmichel, also found by WatchPug

Note that the PoolTemplate.initialize function, called when creating a market with Factory.createMarket, calls a vault function to transfer an initial deposit amount (conditions[1]) from the initial depositor (_references[4]):

// PoolTemplate
function initialize(
     string calldata _metaData,
     uint256[] calldata _conditions,
     address[] calldata _references
) external override {
     // ...

     if (_conditions[1] > 0) {
          // @audit vault calls asset.transferFrom(_references[4], vault, _conditions[1])
          _depositFrom(_conditions[1], _references[4]);
     }
}

function _depositFrom(uint256 _amount, address _from)
     internal
     returns (uint256 _mintAmount)
{
     require(
          marketStatus == MarketStatus.Trading && paused == false,
          "ERROR: DEPOSIT_DISABLED"
     );
     require(_amount > 0, "ERROR: DEPOSIT_ZERO");

     _mintAmount = worth(_amount);
     // @audit vault calls asset.transferFrom(_from, vault, _amount)
     vault.addValue(_amount, _from, address(this));

     emit Deposit(_from, _amount, _mintAmount);

     //mint iToken
     _mint(_from, _mintAmount);
}

The initial depositor needs to first approve the vault contract for the transferFrom to succeed.

An attacker can then frontrun the Factory.createMarket transaction with their own market creation (it does not have access restrictions) and create a market with different parameters but still passing in _conditions[1]=amount and _references[4]=victim.

A market with parameters that the initial depositor did not want (different underlying, old whitelisted registry/parameter contract, etc.) can be created with their tokens and these tokens are essentially lost.

Can the initial depositor be set to Factory.createMarket’s msg.sender, instead of being able to pick a whitelisted one as _references[4]?

oishun1112 (Insure) confirmed:

https://github.com/code-423n4/2022-01-insure-findings/issues/224

[H-05] backdoor in withdrawRedundant

Submitted by cmichel, also found by camden, WatchPug, and Ruhum

The Vault.withdrawRedundant has wrong logic that allows the admins to steal the underlying vault token.

function withdrawRedundant(address _token, address _to)
     external
     override
     onlyOwner
{
     if (
          _token == address(token) &&
          balance < IERC20(token).balanceOf(address(this))
     ) {
          uint256 _redundant = IERC20(token).balanceOf(address(this)) -
               balance;
          IERC20(token).safeTransfer(_to, _redundant);
     } else if (IERC20(_token).balanceOf(address(this)) > 0) {
          // @audit they can rug users. let's say balance == IERC20(token).balanceOf(address(this)) => first if false => transfers out everything
          IERC20(_token).safeTransfer(
               _to,
               IERC20(_token).balanceOf(address(this))
          );
     }
}
POC
  • Vault deposits increase as Vault.addValue is called and the balance increases by _amount as well as the actual IERC20(token).balanceOf(this). Note that balance == IERC20(token).balanceOf(this)
  • Admins call vault.withdrawRedundant(vault.token(), attacker) which goes into the else if branch due to the balance inequality condition being false. It will transfer out all vault.token() amounts to the attacker.

Impact

There’s a backdoor in the withdrawRedundant that allows admins to steal all user deposits.

I think the devs wanted this logic from the code instead:

function withdrawRedundant(address _token, address _to)
     external
     override
     onlyOwner
{
     if (
          _token == address(token)
     ) {
          if (balance < IERC20(token).balanceOf(address(this))) {
               uint256 _redundant = IERC20(token).balanceOf(address(this)) -
                    balance;
               IERC20(token).safeTransfer(_to, _redundant);
          }
     } else if (IERC20(_token).balanceOf(address(this)) > 0) {
          IERC20(_token).safeTransfer(
               _to,
               IERC20(_token).balanceOf(address(this))
          );
     }
}

oishun1112 (Insure) confirmed:

similar to PVE03 (Peckshield audit) We will create a PR and merge after we merge both audit/code4rena and audit/peckshield branches in the InsureDAO repository.

[H-06] the first depositor to a pool can drain all users

Submitted by danb

https://github.com/code-423n4/2022-01-insure/blob/main/contracts/PoolTemplate.sol#L807 if there is no liquidity in the pool, the first deposit determines the total liquidity, if the amount is too small the minted liquidity for the next liquidity providers will round down to zero.

Impact

An attacker can steal all money from liquidity providers.

Proof of Concept

consider the following scenario: a pool is created. the attacker is the first one to deposit, they deposit with _amount == 1, the smallest amount possible. meaning the total liquidity is 1. then they join another pool in order to get attributions in the vault. they transfer the attributions to the pool using transferAttribution. for example, they transferred 1M dollar worth of attributions. the next person deposits in the index, for example, 500,000 dollars. https://github.com/code-423n4/2022-01-insure/blob/main/contracts/PoolTemplate.sol#L803 the amount they will get is:

_amount = (_value * _supply) / _originalLiquidity;

as we know: _amount = 500,000 dollar _supply = 1 _totalLiquidity = 1,000,000 dollar (the attacker transferred directly) the investor will get (500,000 dollar * 1) / (1,000,000 dollar) = 0 and they will pay 500,000 this money will go to the index, and the attacker holds all of the shares, so they can withdraw it and get 1,500,000 stealing 500,000 dollars from the second investor.

oishun1112 (Insure) acknowledged and disagreed with severity:

yes. Every address that has attributions can call transferAttribution(), however, the address has to call addValue() to earn attributions. addValue() has onlyMarket modifier. To pass onlyMarket modifier, ownership has to be stolen, in short. Since we assume ownership control is driven safely, we don’t take this as an issue.

0xean (judge) commented:

Agree with warden that the privilege addresses should not be able to use approvals in a way that rugs users funds.

Based on the fact that we have seen many rug pulls in the space based on compromised “owner” keys, this is a valid attack path.

3 — High: Assets can be stolen/lost/compromised directly (or indirectly if there is a valid attack path that does not have hand-wavy hypotheticals).

[H-07] Wrong design/implementation of permission control allows malicious/compromised Registry or Factory admin to steal funds from users’ wallet balances

Submitted by WatchPug

The current design/implementation allows a market address (registered on registry) to call Vault#addValue() and transfer tokens from an arbitrary address to a specified _beneficiary up the approved amount at any time, and the _beneficiary can withdraw the funds by calling Vault#withdrawAllAttribution() immediately.

This poses a very dangerous risk to all the users that approved their tokens to the Vault contracts (each one holds all users’ allowances for that token).

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Vault.sol#L52-L58

modifier onlyMarket() {
    require(
        IRegistry(registry).isListed(msg.sender),
        "ERROR_ONLY_MARKET"
    );
    _;
}

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Vault.sol#L124-L140

function addValue(
    uint256 _amount,
    address _from,
    address _beneficiary
) external override onlyMarket returns (uint256 _attributions) {

    if (totalAttributions == 0) {
        _attributions = _amount;
    } else {
        uint256 _pool = valueAll();
        _attributions = (_amount * totalAttributions) / _pool;
    }
    IERC20(token).safeTransferFrom(_from, address(this), _amount);
    balance += _amount;
    totalAttributions += _attributions;
    attributions[_beneficiary] += _attributions;
}

Registry owner can call Registry#supportMarket() and mark an arbitrary address as a market.

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Registry.sol#L49-L60

function supportMarket(address _market) external override {
    require(!markets[_market], "ERROR: ALREADY_REGISTERED");
    require(
        msg.sender == factory || msg.sender == ownership.owner(),
        "ERROR: UNAUTHORIZED_CALLER"
    );
    require(_market != address(0), "ERROR: ZERO_ADDRESS");

    allMarkets.push(_market);
    markets[_market] = true;
    emit NewMarketRegistered(_market);
}

Or, the owner of the Factory can call createMarket() to add a malicous market contract via a custom template contract to the markets list.

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Factory.sol#L214-L216

Proof of Concept

A malicious/compromised Registry owner can:

  1. Call Registry#supportMarket() and set markets[attackerAddress] to true;
  2. Call Vault#addValue(token.balanceOf(victimAddress), victimAddress, attackerAddress) and transferring all the balanceOf victim’s wallet to the vault, owned by attackerAddress.
  3. Call Vault#withdrawAllAttribution(attackerAddress) and retrive the funds.

The malicious/compromised Registry owner can repeat the steps above for all the users who approved the Vault contract for all the Vault contracts.

As a result, the attacker can steal all the wallet balances of the tokens approved to the protocol.

Root Cause

Improper access control for using users’ allowances.

Recommendation

Consider changing the design/implementation to make sure that the allowances approved by the users can only be used by themselves.

oishun1112 (Insure) acknowledged and disagreed with severity:

this is an issue only when ownership control has fail. This architecture is necessary to achieve simplicity of the code. We assume ownership control works fine.

0xean (judge) commented:

Agree with warden that the privilege addresses should not be able to use approvals in a way that rugs users funds.

Based on the fact that we have seen many rug pulls in the space based on compromised “owner” keys, this is a valid attack path.

3 — High: Assets can be stolen/lost/compromised directly (or indirectly if there is a valid attack path that does not have hand-wavy hypotheticals).

[H-08] IndexTemplate.sol#compensate() will most certainly fail

Submitted by WatchPug

Precision loss while converting between the amount of shares and the amount of underlying tokens back and forth is not handled properly.

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/IndexTemplate.sol#L438-L447

uint256 _shortage;
if (totalLiquidity() < _amount) {
    //Insolvency case
    _shortage = _amount - _value;
    uint256 _cds = ICDSTemplate(registry.getCDS(address(this)))
        .compensate(_shortage);
    _compensated = _value + _cds;
}
vault.offsetDebt(_compensated, msg.sender);

In the current implementation, when someone tries to resume the market after a pending period ends by calling PoolTemplate.sol#resume(), IndexTemplate.sol#compensate() will be called internally to make a payout. If the index pool is unable to cover the compensation, the CDS pool will then be used to cover the shortage.

However, while CDSTemplate.sol#compensate() takes a parameter for the amount of underlying tokens, it uses vault.transferValue() to transfer corresponding _attributions (shares) instead of underlying tokens.

Due to precision loss, the _attributions transferred in the terms of underlying tokens will most certainly be less than the shortage.

At L444, the contract believes that it’s been compensated for _value + _cds, which is lower than the actual value, due to precision loss.

At L446, when it calls vault.offsetDebt(_compensated, msg.sender), the tx will revert at require(underlyingValue(msg.sender) >= _amount).

As a result, resume() can not be done, and the debt can’t be repaid.

Proof of Concept

Given:

  • vault.underlyingValue = 10,000
  • vault.valueAll = 30,000
  • totalAttributions = 2,000,000
  • _amount = 1,010,000
  • _shortage = _amount - vault.underlyingValue = 1,000,000
  • _attributions = (_amount * totalAttributions) / valueAll = 67,333,333
  • actualValueTransfered = (valueAll * _attributions) / totalAttributions = 1009999

Expected results: actualValueTransfered = _shortage;

Actual results: actualValueTransfered < _shortage.

Impact

The precision loss isn’t just happening on special numbers, but will most certainly always revert the txs.

This will malfunction the contract as the index pool can not compensate(), therefore the pool can not resume(). Causing the funds of the LPs of the pool and the index pool to be frozen, and other stakeholders of the same vault will suffer fund loss from an unfair share of the funds compensated before.

Recommendation

Change to:

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/IndexTemplate.sol#L439-L446

if (totalLiquidity() < _amount) {
    //Insolvency case
    _shortage = _amount - _value;
    uint256 _cds = ICDSTemplate(registry.getCDS(address(this)))
        .compensate(_shortage);
    _compensated = vault.underlyingValue(address(this));
}
vault.offsetDebt(_compensated, msg.sender);

oishun1112 (Insure) confirmed and disagreed with severity

oishun1112 (Insure) resolved

[H-09] Vault#setController() owner of the Vault contracts can drain funds from the Vault

Submitted by WatchPug

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Vault.sol#L485-L496

function setController(address _controller) public override onlyOwner {
    require(_controller != address(0), "ERROR_ZERO_ADDRESS");

    if (address(controller) != address(0)) {
        controller.migrate(address(_controller));
        controller = IController(_controller);
    } else {
        controller = IController(_controller);
    }

    emit ControllerSet(_controller);
}

The owner of the Vault contract can set an arbitrary address as the controller.

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Vault.sol#L342-L352

function utilize() external override returns (uint256 _amount) {
    if (keeper != address(0)) {
        require(msg.sender == keeper, "ERROR_NOT_KEEPER");
    }
    _amount = available(); //balance
    if (_amount > 0) {
        IERC20(token).safeTransfer(address(controller), _amount);
        balance -= _amount;
        controller.earn(address(token), _amount);
    }
}

A malicious controller contract can transfer funds from the Vault to the attacker.

Proof of Concept

A malicious/compromised can:

  1. Call Vault#setController() and set controller to a malicious contract;

    • L489 the old controller will transfer funds to the new, malicious controller.
  2. Call Vault#utilize() to deposit all the balance in the Vault contract into the malicious controller contract.
  3. Withdraw all the funds from the malicious controller contract.

Recommendation

Consider disallowing Vault#setController() to set a new address if a controller is existing, which terminates the possibility of migrating funds to a specified address provided by the owner. Or, putting a timelock to this function at least.

oishun1112 (Insure) acknowledged and disagreed with severity:

we assume ownership control is driven safely

0xean (judge) commented:

Agree with warden that the privilege addresses should not be able to use approvals in a way that rugs users funds.

Based on the fact that we have seen many rug pulls in the space based on compromised “owner” keys, this is a valid attack path. 3 — High: Assets can be stolen/lost/compromised directly (or indirectly if there is a valid attack path that does not have hand-wavy hypotheticals).

[H-10] A malicious/compromised Registry or Factory admin can drain all the funds from the Vault contracts

Submitted by WatchPug

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Vault.sol#L52-L58

modifier onlyMarket() {
    require(
        IRegistry(registry).isListed(msg.sender),
        "ERROR_ONLY_MARKET"
    );
    _;
}

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Vault.sol#L201-L206

function borrowValue(uint256 _amount, address _to) external onlyMarket override {
    debts[msg.sender] += _amount;
    totalDebt += _amount;

    IERC20(token).safeTransfer(_to, _amount);
}

The current design/implementation allows a market address (registered on the registry) to call Vault#borrowValue() and transfer tokens to an arbitrary address.

Proof of Concept

See the PoC section on [WP-H24].

Recommendation

  1. Consider adding constrains (eg. timelock) to Registry#supportMarket().
  2. Consdier adding constrains (upper bound for each pool, and index pool for example) to Vault#borrowValue().

oishun1112 (Insure) acknowledged and disagreed with severity:

Ownership has to be stolen to drain funds using this method and we assume ownership control driven safely, so we don’t treat this as issue

0xean (judge) commented:

Agree with warden that the privilege addresses should not be able to use approvals in a way that rugs users funds.

Based on the fact that we have seen many rug pulls in the space based on compromised “owner” keys, this is a valid attack path. 3 — High: Assets can be stolen/lost/compromised directly (or indirectly if there is a valid attack path that does not have hand-wavy hypotheticals).

[H-11] PoolTemplate.sol#resume() Wrong implementation of resume() will compensate overmuch redeem amount from index pools

Submitted by WatchPug, also found by danb

Wrong arithmetic.

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/PoolTemplate.sol#L700-L717

uint256 _deductionFromIndex = (_debt * _totalCredit * MAGIC_SCALE_1E6) /
            totalLiquidity();
    uint256 _actualDeduction;
    for (uint256 i = 0; i < indexList.length; i++) {
        address _index = indexList[i];
        uint256 _credit = indicies[_index].credit;
        if (_credit > 0) {
            uint256 _shareOfIndex = (_credit * MAGIC_SCALE_1E6) /
                _totalCredit;
            uint256 _redeemAmount = _divCeil(
                _deductionFromIndex,
                _shareOfIndex
            );
            _actualDeduction += IIndexTemplate(_index).compensate(
                _redeemAmount
            );
        }
    }

Proof of Concept

  • totalLiquidity = 200,000* 10**18;
  • totalCredit = 100,000 * 10**18;
  • debt = 10,000 * 10**18;
  • [Index Pool 1] Credit = 20,000 * 10**18;
  • [Index Pool 2] Credit = 30,000 * 10**18;
uint256 _deductionFromIndex = (_debt * _totalCredit * MAGIC_SCALE_1E6) /
            totalLiquidity();
// _deductionFromIndex = 10,000 * 10**6 * 10**18;

[Index Pool 1]:

uint256 _shareOfIndex = (_credit * MAGIC_SCALE_1E6) / _totalCredit;  
//  _shareOfIndex = 200000

uint256 _redeemAmount = _divCeil(
    _deductionFromIndex,
    _shareOfIndex
);

// _redeemAmount = 25,000 * 10**18;

[Index Pool 2]:

uint256 _shareOfIndex = (_credit * MAGIC_SCALE_1E6) / _totalCredit;  
//  _shareOfIndex = 300000

uint256 _redeemAmount = _divCeil(
    _deductionFromIndex,
    _shareOfIndex
);

// _redeemAmount = 16666666666666666666667 (~ 16,666 * 10**18)

In most cases, the transaction will revet on underflow at:

uint256 _shortage = _deductionFromIndex /
    MAGIC_SCALE_1E6 -
    _actualDeduction;

In some cases, specific pools will be liable for unfair compensation:

If the CSD is empty, Index Pool 1 only have 6,000 * 10**18 and Index Pool 2 only have 4,000 * 10**18, the _actualDeduction will be 10,000 * 10**18, _deductionFromPool will be 0.

Index Pool 1 should only pay 1,000 * 10**18, but actually paid 6,000 * 10**18, the LPs of Index Pool 1 now suffer funds loss.

Recommendation

Change to:

uint256 _deductionFromIndex = (_debt * _totalCredit * MAGIC_SCALE_1E6) / totalLiquidity();
uint256 _actualDeduction;
for (uint256 i = 0; i < indexList.length; i++) {
    address _index = indexList[i];
    uint256 _credit = indicies[_index].credit;
    if (_credit > 0) {
        uint256 _shareOfIndex = (_credit * MAGIC_SCALE_1E6) /
            _totalCredit;
        uint256 _redeemAmount = _divCeil(
            _deductionFromIndex * _shareOfIndex,
            MAGIC_SCALE_1E6 * MAGIC_SCALE_1E6
        );
        _actualDeduction += IIndexTemplate(_index).compensate(
            _redeemAmount
        );
    }
}

oishun1112 (Insure) confirmed and resolved

[H-12] IndexTemplate.sol Wrong implementation allows lp of the index pool to resume a locked PayingOut pool and escape the responsibility for the compensation

Submitted by WatchPug, also found by leastwood

Based on the context, the system intends to lock all the lps during PayingOut period.

However, the current implementation allows anyone, including LPs to call resume() and unlock the index pool.

It allows a malicious LP to escape the responsibility for the compensation, at the expense of other LPs paying more than expected.

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/IndexTemplate.sol#L459-L471

function resume() external override {
    uint256 _poolLength = poolList.length;

    for (uint256 i = 0; i < _poolLength; i++) {
        require(
            IPoolTemplate(poolList[i]).paused() == false,
            "ERROR: POOL_IS_PAUSED"
        );
    }

    locked = false;
    emit Resumed();
}

Recommendation

Change to:

function resume() external override {
   uint256 _poolLength = poolList.length;

   for (uint256 i = 0; i < _poolLength; i++) {
       require(
           IPoolTemplate(poolList[i]).marketStatus() == MarketStatus.Trading,
           "ERROR: POOL_IS_PAYINGOUT"
       );
   }

   locked = false;
   emit Resumed();
}

oishun1112 (Insure) confirmed

[H-13] Admin of the index pool can withdrawCredit() after applyCover() to avoid taking loss for the compensation paid for a certain pool

Submitted by WatchPug

In the current implementation, when an incident is reported for a certain pool, the index pool can still withdrawCredit() from the pool, which in the best interest of an index pool, the admin of the index pool is preferred to do so.

This allows the index pool to escape from the responsibility for the risks of invested pools.

Making the LPs of the pool take an unfair share of the responsibility.

Proof of Concept
  • Pool A totalCredit = 10,000
  • Pool A rewardPerCredit = 1
  • [Index Pool 1] allocates 1,000 credits to Pool A:
  • totalCredit = 11,000
  • indicies[Index Pool 1] = 1,000
  • After a while, Pool A rewardPerCredit has grown to 1.1, and applyCover() has been called, [Index Pool 1] call withdrawCredit() get 100 premium
  • totalCredit = 10,000
  • indicies[Index Pool 1] = 0
  • After pendingEnd, the pool resume(),[ Index Pool 1] will not be paying for the compensation since credit is 0.

In our case, [Index Pool 1] earned premium without paying for a part of the compensation.

Recommendation

Change to:

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/PoolTemplate.sol#L416-L421

function withdrawCredit(uint256 _credit)
    external
    override
    returns (uint256 _pending)
{
    require(
        marketStatus == MarketStatus.Trading,
        "ERROR: WITHDRAW_CREDIT_BAD_CONDITIONS"
    );
    IndexInfo storage _index = indicies[msg.sender];

oishun1112 (Insure) confirmed and disagreed with severity:

to call PoolTemplate: withdrawCredit(), someone has to call IndexTemplate: withdraw(), set(), and adjustAlloc().

set() is onlyOwner, so we assume it’s fine() adjustAlloc() is public. this clean up and flatten the credit distribution. withdraw() is public. this reduce totalCredit to distribute. when exceed upperSlack, call adjustAlloc().

We should lock the credit control when pool is in payout status. This implementation, still allows small amount of withdraw, for users who were requested Withdraw.

oishun1112 (Insure) commented:

We have fixed with PVE02 (Peckshield audit) issue together.

Medium Risk Findings (9)

[M-01] repayDebt in Vault.sol could DOS functionality for markets

Submitted by p4st13r4

Any user can pay the debt for any borrower in Vault.sol, by using repayDebt(). This function allows anyone to repay any amount of borrowed value, up-to and including the totalDebt value; it works by setting the debts[_target] to zero, and decreasing totalDebt by the given amount, up to zero. However, all debts of the other borrowers are left untouched.

If a malicious (but generous) user were to repay the debt for all the borrowers, markets functionality regarding borrowing would be DOSed: the vault would try to decrease the debt of the market, successfully, but would fail to decrease totalDebt as it would result in an underflow

Proof of Concept

https://github.com/code-423n4/2022-01-insure/blob/main/contracts/Vault.sol#L257

Make repayDebt() accept an amount up-to and including the value of the debt for the given borrower

oishun1112 (Insure) confirmed:

this needs to be specified how in more detail.

[M-02] Owner can call applyCover multiple times in PoolTemplate.sol

Submitted by camden

The owner could potentially extend the insurance period indefinitely in the applyCover function without ever allowing the market to resume. This is because there is no check in applyCover to ensure that the market is in a Trading state.

This can also allow the owner to emit fraudulent MarketStatusChanged events.

Require that the market be in a Trading state to allow another applyCover call.

oishun1112 (Insure) confirmed and resolved:

this behaviour is not intended, so confirmed.

0xean (judge) commented:

upgrading to medium severity since it alters the function of the protocol

2 — Med: Assets 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.

[M-03] Signature replay

Submitted by 0x1f8b

Signature replay in PoolTemplate.

Proof of Concept

The redeem method of PoolTemplate verifies the data stored in incident, and the verification logic of this process is performed as following:

require(
    MerkleProof.verify(
        _merkleProof,
        _targets,
        keccak256(
            abi.encodePacked(_insurance.target, _insurance.insured)
        )
    ) ||
        MerkleProof.verify(
            _merkleProof,
            _targets,
            keccak256(abi.encodePacked(_insurance.target, address(0)))
        ),
    "ERROR: INSURANCE_EXEMPTED"
);

As can be seen, the only data related to the _insurance aretarget and insured, so as the incident has no relation with theInsurance, apparently nothing prevents a user to call insure with high amounts, after receive the incident, the only thing that prevents this from being reused is that the owner creates the incident with an _incidentTimestamp from the past.

So if an owner create a incident from the future it’s possible to create a new insure that could be reused by the same affected address.

Another lack of input verification that could facilitate this attack is the _span=0 in the insure method.

It is mandatory to add a check in applyCover that_incidentTimestamp is less than the current date and the span argument is greater than 0 in theinsure method.

oishun1112 (Insure) confirmed and resolved, but disagreed with severity:

agree on the _incidentTimestamp check. disagree on span check since there already is

require(
            parameters.getMinDate(msg.sender) <= _span,
            "ERROR: INSURE_SPAN_BELOW_MIN"
        );

we are going to set default value of 1week for everyone

oishun1112 (Insure) commented:

we assume ownership control works fine. this can lose money in-proper way, but not at risk since onlyOwner modifier applied.

0xean (judge) commented:

going to leave this as 2

2 — Med: Assets 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.

The external requirement here would be an incorrect timestamp from the owner which would cause assets to be at risk from the replay.

oishun1112 (Insure) commented:

there is

_span >= getMinDate()

so we don’t implement _span > 0

[M-04] System Debt Is Not Handled When Insurance Pools Become Insolvent

Submitted by leastwood

If an incident has occurred where an insurance policy is to be redeemed. The market is put into the MarketStatus.Payingout mode where the _insurance.insured account is allowed to redeem their cover and receive a payout amount. Upon paying out the insurance cover, any user is able to resume the market by calling PoolTemplate.resume(). This function will compensate the insurance pool if it is insolvent by querying IndexTemplate.compensate() which in turn queries CDSTemplate.compensate() to cover any shortage.

In the event none of these entities are able to cover the shortage in debt, the system accrues the debt. However, there is currently no mechanism to ensure when transferDebt() is called in PoolTemplate.resume(), the accrued system debt is paid off. Therefore, the system may incorrectly handle insolvency on an extreme edge case, generating system instability.

Proof of Concept

https://github.com/code-423n4/2022-01-insure/blob/main/contracts/PoolTemplate.sol#L691-L734

function resume() external {
    require(
        marketStatus == MarketStatus.Payingout &&
            pendingEnd < block.timestamp,
        "ERROR: UNABLE_TO_RESUME"
    );

    uint256 _debt = vault.debts(address(this));
    uint256 _totalCredit = totalCredit;
    uint256 _deductionFromIndex = (_debt * _totalCredit * MAGIC_SCALE_1E6) /
        totalLiquidity();
    uint256 _actualDeduction;
    for (uint256 i = 0; i < indexList.length; i++) {
        address _index = indexList[i];
        uint256 _credit = indicies[_index].credit;
        if (_credit > 0) {
            uint256 _shareOfIndex = (_credit * MAGIC_SCALE_1E6) /
                _totalCredit;
            uint256 _redeemAmount = _divCeil(
                _deductionFromIndex,
                _shareOfIndex
            );
            _actualDeduction += IIndexTemplate(_index).compensate(
                _redeemAmount
            );
        }
    }

    uint256 _deductionFromPool = _debt -
        _deductionFromIndex /
        MAGIC_SCALE_1E6;
    uint256 _shortage = _deductionFromIndex /
        MAGIC_SCALE_1E6 -
        _actualDeduction;

    if (_deductionFromPool > 0) {
        vault.offsetDebt(_deductionFromPool, address(this));
    }

    vault.transferDebt(_shortage);

    marketStatus = MarketStatus.Trading;
    emit MarketStatusChanged(MarketStatus.Trading);
}

Consider devising a mechanism to ensure system debt is properly handled. After discussions with the sponsor, it seems that they will be implementing a way to mint INSURE tokens which will be used to cover the shortfall.

oishun1112 (Insure) acknowledged

yes, PoolTemplate calls transferDebt() to make his debt to the system debt in case all Index and CDS layers couldn’t cover the shortage. In this case, we have to repay the system debt somehow since this is the situation that we over-lose money. One way is that someone calls repayDebt() and pay for it (not realistic at all). As we implement the way to payback, we are considering minting INSURE token or, other better mechanism.

This is not developed yet, and acknowledged.

[M-05] Vault.sol Tokens with fee on transfer are not supported

Submitted by WatchPug, also found by pmerkleplant, cmichel, Ruhum, and Dravee

There are ERC20 tokens that charge fee for every transfer() / transferFrom().

Vault.sol#addValue() assumes that the received amount is the same as the transfer amount, and uses it to calculate attributions, balance amounts, etc. While the actual transferred amount can be lower for those tokens.

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/Vault.sol#L124-L140

function addValue(
    uint256 _amount,
    address _from,
    address _beneficiary
) external override onlyMarket returns (uint256 _attributions) {

    if (totalAttributions == 0) {
        _attributions = _amount;
    } else {
        uint256 _pool = valueAll();
        _attributions = (_amount * totalAttributions) / _pool;
    }
    IERC20(token).safeTransferFrom(_from, address(this), _amount);
    balance += _amount;
    totalAttributions += _attributions;
    attributions[_beneficiary] += _attributions;
}

Recommendation

Consider comparing before and after balance to get the actual transferred amount.

oishun1112 (Insure) acknowledged:

only USDC can be underwriting asset. We are trying to make it multi currency. We won’t implement this now due to the gas consumption, but we will when we develop new type of Vault

[M-06] Index compensate is 0 when totalLiquidity() is enough to cover the whole amount

Submitted by pauliax

In IndexTemplate, function compensate, When \_amount > \_value, and <= totalLiquidity(), the value of \_compensated is not set, so it gets a default value of 0:

if (_value >= _amount) {
    ...
    _compensated = _amount;
} else {
    ...
    if (totalLiquidity() < _amount) {
        ...
        _compensated = _value + _cds;
    }
    vault.offsetDebt(_compensated, msg.sender);
}

But nevertheless, in both cases, it calls vault.offsetDebt, even when the\_compensated is 0 (no else block).

I think, in this case, it should try to redeem the premium (withdrawCredit?) to cover the whole amount, but I am not sure about the intentions as I didn’t have enough time to understand this protocol in depth.

oishun1112 (Insure) confirmed and resolved:

Right. totalLiquidity = underlyingValue + pendingPremium.

I will discuss how to fix this issue with my team

[M-07] requestWithdraw without obligation to withdraw allow underwriter to avoid payout

Submitted by gzeon

To prevent withdrawal front-running, a lockup period is set between withdrawal request and withdrawal. However, there are no obligation to withdraw after the lockup period and the capital will keep earning premium during lockup. A strategy for underwriter is to keep requesting withdrawal every lockup period to keep their average lockup to lockup period/2.

Proof of Concept

https://github.com/code-423n4/2022-01-insure/blob/19d1a7819fe7ce795e6d4814e7ddf8b8e1323df3/contracts/PoolTemplate.sol#L279

Assuming

  1. Reporting DAO vote last for 24 hours (according to docs) plus there will be delay between the hack and vote creation
  2. the lockup period is set to 86400 (24 hours) in the supplied test cases

It is very likely an underwriter can avoid payout by such strategy since their effective lockup would be 12 hours only. They will continue to earn yield in the pool and only require some extra gas cost for the requestWithdraw every 24 hours.

Extend the lockup period at least by a factor of 2 or force underwriter to withdraw after lockup period.

oishun1112 (Insure) acknowledged:

Yes, lock up period is going to be like a week~2week in production.

[M-08] Unbounded iteration over all indexes (2)

Submitted by Dravee, also found by robee, egjlmn1, danb, WatchPug, Fitraldys, and Ruhum

The transactions could fail if the array get too big and the transaction would consume more gas than the block limit. This will then result in a denial of service for the desired functionality and break core functionality.

Proof of Concept

https://github.com/code-423n4/2022-01-insure/blob/main/contracts/PoolTemplate.sol#L703

Tools Used

VS Code

Keep the array size small.

oishun1112 (Insure) confirmed

0xean (judge) commented:

Upgrading to sev-2 as this will eventually affect the availability of the protocol as transactions revert.

Low Risk Findings (42)

Non-Critical Findings (15)

Gas Optimizations (98)

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.