Security analysis tool for EVM bytecode. Supports smart contracts built for Ethereum, Hedera, Quorum, Vechain, Roostock, Tron and other EVM-compatible blockchains.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
mythril/README.md

6.6 KiB

Mythril

Mythril is a reverse engineering and bug hunting framework for the Ethereum blockchain.

Installation and setup

Install from Pypi:

$ pip install mythril

Or, clone the GitHub repo to install the newest master branch:

$ git clone https://github.com/b-mueller/mythril/
$ cd mythril
$ python setup.py install

You also need a go-ethereum node that is synced with the network (note that Mythril uses non-standard RPC APIs only supported by go-ethereum, so other clients likely won't work). Start the node as follows:

$ geth --rpc --rpcapi eth,admin,debug --syncmode fast

Database initialization

Mythril builds its own contract database to enable fast search operations. This is to enable operations like those described in the legendary "Mitch Brenner" blog post in seconds instead of days. Unfortunately, the initial sync process is slow. You don't need to sync the whole blockchain right away though: If you abort the syncing process with ctrl+c, it will be auto-resumed the next time you run the --init-db command.

$ myth --init-db
Starting synchronization from latest block: 4323706
Processing block 4323000, 3 individual contracts in database
(...)

Note that syncing doesn't take quite as long as it first seems, because the blocks get smaller towards the beginning of the chain.

The default behavior is to only sync contracts with a non-zero balance. You can disable this behavior with the --sync-all flag, but be aware that this will result in a huge (as in: dozens of GB) database.

Command line usage

The Mythril command line tool (aptly named myth) allows you to conveniently access some of Mythril's functionality.

Searching the database

The search feature allows you to find contract instances that contain specific function calls and opcode sequences. It supports simple boolean expressions, such as:

$ myth --search "func#changeMultisig(address)#"
$ myth --search "code#PUSH1 0x50,POP#"
$ myth --search "func#changeMultisig(address)# and code#PUSH1 0x50#"

Disassembler

Use the -d flag to disassemble code. The disassembler accepts a bytecode string or a contract address as its input.

$ myth -d -c "$ ./myth -d -c "5060"
0 PUSH1 0x60

Specifying an address via -a ADDRESS will download the contract code from your node. Mythril will try to resolve function names using the signatures in database/signature.json:

$ myth -d -a "0x2a0c0dbecc7e4d658f48e01e3fa353f44050c208"
0 PUSH1 0x60
2 PUSH1 0x40
4 MSTORE
(...)
1135 - FUNCTION safeAdd(uint256,uint256) -
1136 CALLVALUE
1137 ISZERO

Adding the -g FILENAME option will output a call graph:

$ myth -d -a "0xFa52274DD61E1643d2205169732f29114BC240b3" -g ./graph.svg

callgraph

Note that currently, Mythril only processes JUMP and JUMPI instructions with immediately preceding PUSH, but doesn't understand dynamic jumps and function calls.

Tracing Code

You can run a code trace in the PyEthereum virtual machine. Optionally, input data can be passed via the --data flag.

$ myth -t -a "0x3665f2bf19ee5e207645f3e635bf0f4961d661c0"
vm storage={'storage': {}, 'nonce': '0', 'balance': '0', 'code': '0x'} gas=b'21000' stack=[] address=b'6e\xf2\xbf\x19\xee^ vE\xf3\xe65\xbf\x0fIa\xd6a\xc0' depth=0 steps=0 inst=96 pushvalue=96 pc=b'0' op=PUSH1
vm op=PUSH1 gas=b'20997' stack=[b'96'] depth=0 steps=1 inst=96 pushvalue=64 pc=b'2'
vm op=MSTORE gas=b'20994' stack=[b'96', b'64'] depth=0 steps=2 inst=82 pc=b'4'

Finding cross-references

It is often useful to find other contracts referenced by a particular contract. Let's assume you want to search for contracts that fulfill conditions similar to the Parity Multisig Wallet Bug. First, you want to find a list of contracts that use the DELEGATECALL opcode:

$ myth --search "code#DELEGATECALL#"
Matched contract with code hash 07459966443977122e639cbf7804c446
Address: 0x76799f77587738bfeef09452df215b63d2cfb08a, balance: 1000000000000000
Address: 0x3582d2a3b67d63ed10f1ecaef0dca71b9283b543, balance: 92000000000000000000
Address: 0x4b9bc00c35f7cee95c65c3c9836040c37dec9772, balance: 89000000000000000000
Address: 0x156d5687a201affb3f1e632dcfb9fde4b0128211, balance: 29500000000000000000
(...)

Note that "code hash" in the above output refers to the contract's index in the database. The following lines ("Address: ...") list instances of same contract deployed on the blockchain.

You can then use the --xrefs flag to find the addresses of referenced contracts:

$ myth --xrefs 07459966443977122e639cbf7804c446
5b9e8728e316bbeb692d22daaab74f6cbf2c4691

Custom scripts

By combining Mythril and PyEthereum modules, you can automate more complex static and dynamic analysis tasks. Here is an example.

Issues

The RPC database sync solution is not very efficient. I explored some other options, including:

  • Using PyEthereum: I encountered issues syncing PyEthereum with Homestead. Also, PyEthApp only supports Python 2.7, which causes issues with other important packages.
  • Accessing the Go-Ethereum LevelDB: This would be a great option. However, PyEthereum database code seems unable to deal with Go-Ethereum's LevelDB. It would take quite a bit of effort to figure this out.
  • IPC might allow for faster sync then RPC - haven't tried it yet.

I'm writing this in my spare time, so contributors would be highly welcome!

Credit

  • JSON RPC library is adapted from ethjsonrpc (it doesn't seem to be maintained anymore, and I needed to make some changes to it).

  • The signature data in signatures.json has been obtained from the Ethereum Function Signature Database.

Disclaimer: Act responsibly!

The purpose of project is to aid discovery of vulnerable smart contracts on the Ethereum mainnet and support research for novel security flaws. If you do find an exploitable issue or vulnerable contract instances, please do the right thing. Also, note that vulnerability branding ("etherbleed", "chainshock",...) is highly discouraged as it will annoy the author and others in the security community.