ERC-20 with Safety Rails
Introduction
One of the great things about Ethereum is that there is no central authority that can modify or undo your transactions. One of the great problems with Ethereum is that there is no central authority with the power to undo user mistakes or illicit transactions. In this article you learn about some of the common mistakes that users commit with ERC-20 tokens, as well as how to create ERC-20 contracts that help users to avoid those mistakes, or that give a central authority some power (for example to freeze accounts).
Note that while we will use the OpenZeppelin ERC-20 token contract(opens in a new tab), this article does not explain it in great details. You can find this information here.
If you want to see the complete source code:
- Open the Remix IDE(opens in a new tab).
- Click the clone github icon ().
- Clone the github repository
https://github.com/qbzzt/20220815-erc20-safety-rails
. - Open contracts > erc20-safety-rails.sol.
Creating an ERC-20 contract
Before we can add the safety rail functionality we need an ERC-20 contract. In this article we'll use the OpenZeppelin Contracts Wizard(opens in a new tab). Open it in another browser and follow these instructions:
Select ERC20.
Enter these settings:
Parameter Value Name SafetyRailsToken Symbol SAFE Premint 1000 Features None Access Control Ownable Upgradability None Scroll up and click Open in Remix (for Remix) or Download to use a different environment. I'm going to assume you're using Remix, if you use something else just make the appropriate changes.
We now have a fully functional ERC-20 contract. You can expand
.deps
>npm
to see the imported code.Compile, deploy, and play with the contract to see that it functions as an ERC-20 contract. If you need to learn how to use Remix, use this tutorial(opens in a new tab).
Common mistakes
The mistakes
Users sometimes send tokens to the wrong address. While we cannot read their minds to know what they meant to do, there are two error types that happen a lot and are easy to detect:
Sending the tokens to the contract's own address. For example, Optimism's OP token(opens in a new tab) managed to accumulate over 120,000(opens in a new tab) OP tokens in less than two months. This represents a significant amount of wealth that presumably people just lost.
Sending the tokens to an empty address, one that doesn't correspond to an externally owned account or a smart contract. While I don't have statistics on how often this happens, one incident could have cost 20,000,000 tokens(opens in a new tab).
Preventing transfers
The OpenZeppelin ERC-20 contract includes a hook, _beforeTokenTransfer
(opens in a new tab), that is called before a token is transferred. By default this hook does not do anything, but we can hang our own functionality on it, such as checks that revert if there's a problem.
To use the hook, add this function after the constructor:
1 function _beforeTokenTransfer(address from, address to, uint256 amount)2 internal virtual3 override(ERC20)4 {5 super._beforeTokenTransfer(from, to, amount);6 }Kopiëren
Some parts of this function may be new if you aren't very familiar with Solidity:
1 internal virtualKopiëren
The virtual
keyword means that just as we inherited functionality from ERC20
and overrode this function, other contracts can inherit from us and override this function.
1 override(ERC20)Kopiëren
We have to specify explicitly that we're overriding(opens in a new tab) the ERC20 token definition of _beforeTokenTransfer
. In general, explicit definitions are a lot better, from the security standpoint, than implicit ones - you cannot forget that you've done something if it's right in front of you. That is also the reason we need to specify which superclass's _beforeTokenTransfer
we are overriding.
1 super._beforeTokenTransfer(from, to, amount);Kopiëren
This line calls the _beforeTokenTransfer
function of the contract or contracts from which we inherited which have it. In this case, that is only ERC20
, Ownable
does not have this hook. Even though currently ERC20._beforeTokenTransfer
doesn't do anything, we call it in case functionality is added in the future (and we then decide to redeploy the contract, because contracts don't change after deployment).
Coding the requirements
We want to add these requirements to the function:
- The
to
address cannot equaladdress(this)
, the address of the ERC-20 contract itself. - The
to
address cannot be empty, it has to be either:- An externally owned account (EOA). We can't check if an address is an EOA directly, but we can check an address's ETH balance. EOAs almost always have a balance, even if they are no longer used - it's difficult to clear them to the last wei.
- A smart contract. Testing if an address is a smart contract is a bit harder. There is an opcode that checks the external code length, called
EXTCODESIZE
(opens in a new tab), but it is not available directly in Solidity. We have to use Yul(opens in a new tab), which is EVM assembly, for it. There are other values we could use from Solidity (<address>.code
and<address>.codehash
(opens in a new tab)), but they cost more.
Lets go over the new code line by line:
1 require(to != address(this), "Can't send tokens to the contract address");Kopiëren
This is the first requirement, check that to
and this(address)
are not the same thing.
1 bool isToContract;2 assembly {3 isToContract := gt(extcodesize(to), 0)4 }Kopiëren
This is how we check if an address is a contract. We cannot receive output directly from Yul, so instead we define a variable to hold the result (isToContract
in this case). The way Yul works is that every opcode is considered a function. So first we call EXTCODESIZE
(opens in a new tab) to get the contract size, and then use GT
(opens in a new tab) to check it is not zero (we are dealing with unsigned integers, so of course it can't be negative). We then write the result to isToContract
.
1 require(to.balance != 0 || isToContract, "Can't send tokens to an empty address");Kopiëren
And finally, we have the actual check for empty addresses.
Administrative access
Sometimes it is useful to have an administrator that can undo mistakes. To reduce the potential for abuse, this administrator can be a multisig(opens in a new tab) so multiple people have to agree on an action. In this article we'll have two administrative features:
Freezing and unfreezing accounts. This can be useful, for example, when an account might be compromised.
Asset cleanup.
Sometimes frauds send fraudulent tokens to the real token's contract to gain legitimacy. For example, see here(opens in a new tab). The legitimate ERC-20 contract is 0x4200....0042(opens in a new tab). The scam that pretends to be it is 0x234....bbe(opens in a new tab).
It is also possible that people send legitimate ERC-20 tokens to our contract by mistake, which is another reason to want to have a way to get them out.
OpenZeppelin provides two mechanisms to enable administrative access:
Ownable
(opens in a new tab) contracts have a single owner. Functions that have theonlyOwner
modifier(opens in a new tab) can only be called by that owner. Owners can transfer ownership to somebody else or renounce it completely. The rights of all other accounts are typically identical.AccessControl
(opens in a new tab) contracts have role based access control (RBAC)(opens in a new tab).
For the sake of simplicity, in this article we use Ownable
.
Freezing and thawing contracts
Freezing and thawing contracts requires several changes:
A mapping(opens in a new tab) from addresses to booleans(opens in a new tab) to keep track of which addresses are frozen. All values are initially zero, which for boolean values is interpreted as false. This is what we want because by default accounts are not frozen.
1 mapping(address => bool) public frozenAccounts;KopiërenEvents(opens in a new tab) to inform anybody interested when an account is frozen or thawed. Technically speaking events are not required for these actions, but it helps off chain code to be able to listen to these events and know what is happening. It's considered good manners for a smart contract to emit them when something that miught be relevant to somebody else happens.
The events are indexed so will be possible to search for all the times an account has been frozen or thawed.
1 // When accounts are frozen or unfrozen2 event AccountFrozen(address indexed _addr);3 event AccountThawed(address indexed _addr);KopiërenFunctions for freezing and thawing accounts. These two functions are nearly identical, so we'll only go over the freeze function.
1 function freezeAccount(address addr)2 public3 onlyOwnerKopiërenFunctions marked
public
(opens in a new tab) can be called from other smart contracts or directly by a transaction.1 {2 require(!frozenAccounts[addr], "Account already frozen");3 frozenAccounts[addr] = true;4 emit AccountFrozen(addr);5 } // freezeAccountKopiërenIf the account is already frozen, revert. Otherwise, freeze it and
emit
an event.Change
_beforeTokenTransfer
to prevent money being moved from a frozen account. Note that money can still be transferred into the frozen account.1 require(!frozenAccounts[from], "The account is frozen");Kopiëren
Asset cleanup
To release ERC-20 tokens held by this contract we need to call a function on the token contract to which they belong, either transfer
(opens in a new tab) or approve
(opens in a new tab). There's no point wasting gas in this case on allowances, we might as well transfer directly.
1 function cleanupERC20(2 address erc20,3 address dest4 )5 public6 onlyOwner7 {8 IERC20 token = IERC20(erc20);Kopiëren
This is the syntax to create an object for a contract when we receive the address. We can do this because we have the definition for ERC20 tokens as part of the source code (see line 4), and that file includes the definition for IERC20(opens in a new tab), the interface for an OpenZeppelin ERC-20 contract.
1 uint balance = token.balanceOf(address(this));2 token.transfer(dest, balance);3 }Kopiëren
This is a cleanup function, so presumably we don't want to leave any tokens. Instead of getting the balance from the user manually, we might as well automate the process.
Conclusion
This is not a perfect solution - there is no perfect solution for the "user made a mistake" problem. However, using these kinds of checks can at least prevent some mistakes. The ability to freeze accounts, while dangerous, can be used to limit the damage of certain hacks by denying the hacker the stolen funds.
Laatst bewerkt: @omahs(opens in a new tab), 17 februari 2024