以太坊安全之EVM与短地址攻击

前言以太坊（Ethereum）是一个开源的有智能合约功能的公共区块链平台，通过其专用加密货币以太币（ETH）提供去中心化的以太坊虚拟机（EVM）来处理点对点合约。

EVM（Ethereum）虚拟机），以太坊虚拟机的简称，以太坊的核心之一。

智能合约的创建和执行都由EVM来完成，简单来说，EVM是一个状态执行的机器，输入是solidity编译后的二进制指令和例程的状态数据，输出是异步状态的改变。

以太坊短地址攻击，最初由Golem团队于2017年4月提出，是由于EVM的设计缺陷导致的突破。

ERC20代币标准定义的转账函数如下：function transfer(address to uint256 value) public returns (bool success)如果前缀的to是末尾前缀的短地址，则EVM在后面字节补足地址，而最后的value值不足则用0填充，导致实际转出的代币数值倍增。

本文从以太坊原始代码的角度分析EVM重叠是如何处理执行智能合约字节码的，并简要分析短地址攻击的原理。

EVM原始分析evm.goEVM源码的位于go-ethereum/core/vm/目录下，在evm.go中定义了EVM结构体，并实现了EVM.Call，EVM.CallCode，EVM.DelegateCall，EVM.StaticCall四种方法来调用智能合约，EVM.Call实现了基本的合约调用的功能，三种后面与方法EVM.Call略有区别，但最终都调用run函数来解析执行智能合约EVM呼叫// Call executes the contract associated with the addr with the given input as// parameters. It also handles any necessary value transfer required and takes// the necessary steps to create accounts and reverses the state in case of an// execution error or failed value transfer.//hunya// 基本的合约调用func (evm *EVM) Call(caller ContractRef addr common.Address input []byte gas uint64 value*big.Int) (ret []byte leftOverGas uint64 err error) { if evm.vmConfig.NoRecursion && evm.depth > 0 { return nil gas nil }// Fail if we're trying to execute above the call depth limitif evm.depth > int(params.CallCreateDepth) { return nil gas ErrDepth } // Fail if we're trying to transfer more than the available balanceif !evm.Context.CanTransfer(evm.StateDB caller.Address() value) { return nil gas ErrInsufficientBalance }var ( to = AccountRef(addr) snapshot = evm.StateDB.Snapshot() )if !evm.StateDB.Exist(addr) { precompiles := PrecompiledContractsHomestead ifevm.chainRules.IsByzantium { precompiles = PrecompiledContractsByzantium } if evm.chainRules.IsIstanbul { precompiles = PrecompiledContractsIstanbul } if precompiles[addr] == nil && evm.chainRules.IsEIP158 && value.Sign() == 0 { // Calling a non existing account don't do anything but ping the tracerif evm.vmConfig.Debug && evm.depth == 0{ evm.vmConfig.Tracer.CaptureStart(caller.Address() addr false input gas value) evm.vmConfig.Tracer.CaptureEnd(ret 0 0 nil) } return nil gas nil }evm.StateDB.CreateAccount(addr) } evm.Transfer(evm.StateDB caller.Address() to.Address() value) // Initialise a new contract and set the code that is to be used by the EVM.// The contract is a scoped environment for this execution context only.contract := NewContract(caller to value gas) contract.SetCallCode(&addrevm.StateDB.GetCodeHash(addr) evm.StateDB.GetCode(addr))// Even if the account has no code we need to continue because it might be a precompilestart := time.Now()// Capture the tracer start/end events in debug mode// debug模式会捕获tracer的start/end事件if evm.vmConfig.Debug && evm.depth == 0{ evm.vmConfig.Tracer.CaptureStart(caller.Address() addr false input gas value)defer func() { // Lazy evaluation of the parametersevm.vmConfig.Tracer.CaptureEnd(ret gas-contract.Gas time.Since(start) err) }() }ret err = run(evm contract input false)//hunya// 调用run函数执行合约// When an error was returned by the EVM or when setting the creation code// above we revert to the snapshot and consume any gas remaining. Additionally// when we're in homestead this also counts for code storage gas errors.if err != nil { evm.StateDB.RevertToSnapshot(snapshot) if err != errExecutionReverted { eGas(contract.Gas) } } return ret contract.Gas err}EVM.CallCode// CallCode executes the contract associated with the addr with the given input// as parameters. It also handles any necessary value transfer required and takes// the necessary steps to create accounts and reverses the state in case of an// execution error or failed value transfer.//// CallCode differs from Call in the sense that it executes the given address'// code with the caller as context.//hunya// 类似solidity中的call函数，调用外部合约，执行上下文在被调用合约中func (evm *EVM) CallCode(caller ContractRef addr common.Address input []byte gas uint64 value *big.Int) (ret []byte leftOverGas uint64 err error) { if evm.vmConfig.NoRecursion && evm.depth > 0 { return nil gas nil }// Fail if we're trying to execute above the call depth limitif evm.depth > int(params.CallCreateDepth) { return nil gas ErrDepth } // Fail if we're trying to transfer more than the available balanceif !evm.CanTransfer(evm.StateDB caller.Address() value) { return nil gas ErrInsufficientBalance }var ( snapshot = evm.StateDB.Snapshot() to = AccountRef(caller.Address()) ) // Initialise a new contract and set the code that is to be used by the EVM.// The contract is a scoped environment for this execution context only.contract := NewContract(caller to value gas) contract.SetCallCode(&addrevm.StateDB.GetCodeHash(addr) evm.StateDB.GetCode(addr))ret err = run(evm contract input false)//hunya// 调用run函数执行合约if err != nil { evm.StateDB.RevertToSnapshot(snapshot) if err != errExecutionReverted { eGas(contract.Gas) } } return ret contract.Gas err}EVM.DelegateCall// DelegateCall executes the contract associated with the addr with the given input// as parameters. It reverses the state in case of an execution error.//// DelegateCall differs from CallCode in the sense that it executes the given address'// code with the caller as context and the caller is set to the caller of the caller.//hunya// 类似solidity中的delegatecall函数，调用外部合约，执行上下文在调用合约中func (evm *EVM) DelegateCall(caller ContractRef addr common.Address input []byte gas uint64) (ret []byte leftOverGas uint64 err error) { if evm.vmConfig.NoRecursion && evm.depth > 0 { return nil gas nil } // Fail if we're trying to execute above the call depth limitif evm.depth > int(params.CallCreateDepth) { return nil gas ErrDepth }var ( snapshot = evm.StateDB.Snapshot() to = AccountRef(caller.Address()) )// Initialise a new contract and make initialise the delegate valuescontract := NewContract(caller to nil gas).AsDelegate() contract.SetCallCode(&addrevm.StateDB.GetCodeHash(addr) evm.StateDB.GetCode(addr))ret err = run(evm contract input false)//hunya// 调用run函数执行合约if err != nil { evm.StateDB.RevertToSnapshot(snapshot) if err != errExecutionReverted { eGas(contract.Gas) } } return ret contract.Gas err}EVM.StaticCall// StaticCall executes the contract associated with the addr with the given input// as parameters while disallowing any modifications to the state during the call.// Opcodes that attempt to perform such modifications will result in exceptions// instead of performing the modifications.//hunya// 与EVM.Call类似，但不允许执行会修改永久存储的数据的指令func (evm *EVM) StaticCall(caller ContractRef addr common.Address input []byte gas uint64) (ret []byte leftOverGas uint64 err error) { if evm.vmConfig.NoRecursion && evm.depth > 0{ return nil gas nil } // Fail if we're trying to execute above the call depth limitif evm.depth > int(params.CallCreateDepth) { return nil gas ErrDepth }var ( to = AccountRef(addr) snapshot = evm.StateDB.Snapshot() ) // Initialise a new contract and set the code that is to be used by the EVM.// The contract is a scoped environment for this execution context only.contract := NewContract(caller to new(big.Int) gas) contract.SetCallCode(&addrevm.StateDB.GetCodeHash(addr) evm.StateDB.GetCode(addr))// We do an AddBalance of zero here just in order to trigger a touch.// This doesn't matter on Mainnet where all empties are gone at the time of Byzantium// but is the correct thing to do and matters on other networks in tests and potential// future scenariosevm.StateDB.AddBalance(addr bigZero)// When an error was returned by the EVM or when setting the creation code// above we revert to the snapshot and consume any gas remaining. Additionally// when we're in Homestead this also counts for code storage gas errors.ret err = run(evm contract input true)//hunya// 调用run函数执行合约if err != nil { evm.StateDB.RevertToSnapshot(snapshot) if err != errExecutionReverted { eGas(contract.Gas) } } return ret contract.Gas err}run函数前半段是判断是否基于太坊内置预编译的特殊合约，有单独的运行方式后半段则是对于一般的合约调用解释器interpreter去执行调用[MISSING IMAGE: , ]interpreter.go解释器相关代码在interpreter.go中，interpreter是一个接口，目前只有EVMInterpreter这一个具体实现[MISSING IMAGE: , ]合约EVM.Call汇率引入Interpreter.Run来到EVMInpreter.RunEVMInterpreter的Run方法代码编码，其中处理执行合约字节码的主循环如下：[MISSING IMAGE: , ]大部分代码主要是检查准备运行环境，执行合约字节码的核心代码主要是以下3行op = contract.GetOp(pc)operation := in.cfg.JumpTable[op]......res err = operation.execute(&pc in contract mem stack)......interpreter的主要工作实际上只是通过JumpTable查找指令，从而实现一个翻译解析的作用最终的执行是通过调用operation对象的execute方法jump_table.gooperation的定义放在jump_table.go中[MISSING IMAGE: , ]jump_table.go中还定义了JumpTable和多种不同的指令集[MISSING IMAGE: , ]在基本指令集中有三个处理input的指令，分别是CALLDATALOAD，CALLDATASIZE和CALLDATACOPY[MISSING IMAGE: , ]jump_table.go中的代码同样只是决议解析的功能，提供了指令的查找，定义了每个指令具体的执行函数Instructions.goinstructions.go中是所有指令的具体实现，上述三个函数的具体实现如下：[MISSING IMAGE: , ]这三个函数的作用分别是从input加载参数入栈，获取input大小，复制input中的参数到内存我们重点关注opCallDataLoad函数是如何处理input中的参数入栈的opCallDataLoad函数调用getDataBig函数，预期contract.Input，stack.pop()和big32，将结果转为big.Int入栈[MISSING IMAGE: , ]getDataBig函数以stack.pop()栈顶元素作为初始索引，截取input中big32大小的数据，然后初始化common.RightPadBytes处理并返回其中涉及到的另外两个函数math.BigMin和common.RightPadBytes如下：//file: go-thereum/common/math/big.gofunc BigMin(x y *big.Int) *big.Int { if x.Cmp(y) > 0 { return y } return x}//file: go-ethereum/common/bytes.gofunc RightPadBytes(slice []byte l int) []byte { if l <= len(slice) { return slice } //右填充0x00至l位padded := make([]byte l) copy(padded slice)return padded}分析到这里，基本上已经能很明显看到问题所在了RightPadBytes会将函数传入的字节切片向左向右填充至l位长度，而l的英文被传入的big32，即32位长度所以在短地址攻击中，调用的transfer(address to uint256 value)函数，如果to是低位更改的地址，由于EVM在处理时是固定截取32位长度的，因此重复value数值高位补的0算进to的末端，而在截取value时由于位数不足32位，则右填充0x00至32位，最终导致转账的value 指数级增长测试与复现编写一个简单的合约来测试pragma solidity ^0.5.0;contract Test { uint256 internal _totalSupply;mapping(address => uint256) internal _balances;event Transfer(address indexed from address indexed to uint256 value);constructor() public { _totalSupply = 1 * 10 ** 18; _balances[msg.sender] =_totalSupply; }function totalSupply() external view returns (uint256) { return _totalSupply; }function balanceOf(address account) external view returns (uint256) { return_balances[account]; }function transfer(address touint256 value) public returns (bool) { require(to != address(0)); require(_balances[msg.sender] >= value); require(_balances[to] + value >= _balances[to]);_balances[msg.sender] -= value; _balances[to] += value; emit Transfer(msg.sender to value); }}混音部署，调用transfer发起正常的转账[MISSING IMAGE: , ]input为0xa9059cbb00000000000000000000000071430fd8c82cc7b991a8455fc6ea5b37a06d393f000000000 0000000000000000000000000000000000000000000000000000001直接尝试短地址攻击，减少去转账地址的后两位，会发现并不能通过，remix会直接报错[MISSING IMAGE: , ]这是因为web3.js做了验证，web3.js是用户与以太坊互连相互的中介原始码复现通过二进制函数复现如下：[MISSING IMAGE: , ]实际复现根据如何完成实际场景的攻击，可以参考文末的链接[1]，利用web3.eth.sendSignedTransaction绕过过限制实际上，web3.js做的校准仅在显着式转移转账地址的函数，如web3.eth.sendTransaction这种，像web3.eth.sendSignedTransaction，web3.eth.sendRawTransaction这种指向的参数是序列化后的数据的就校正不了，是可以完成短地址攻击的，兴趣的可以自己尝试，这里就不多写了PS：文中分析的go-ethereum原始码版本是commit-fdff182，原始码与最新版本有些出入，但最新版本的也未修复这种缺陷（可能官方不认为这是缺陷？），分析思路依然可以沿用思考以太坊曾经EVM并没有修复短地址攻击的这么一个缺陷，而是直接在web3.js里对地址做的校验，目前各种合约或多或少也做了弥补，所以虽然EVM可能可以复现，但实际场景中问题应该不大，但如果是开放RPC的中断可能还是会存在这种风险另外还有一个点，按一下EVM的这种机制，易受攻击的应该换一个transfer(address touint256 value)这个点，只是因为这个函数是ERC20代币标准，而且参数的设计恰好能导致涉及金额的短地址攻击，并且在其他的一些非代币合约，如竞猜，游戏类的合约中，一些非转账类的事务处理函数中，如果不对类似地址这，种的参数做长度校正，可能也存在类似的短地址攻击的风险，也可能并不重复地址，可能还有其他的利用方式还没挖掘出来。