Similarity code detection tool #202
Conversation
|
g seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account. You have signed the CLA already but the status is still pending? Let us recheck it. |
utils/similarity/encode.py
Outdated
| import sys | ||
|
|
||
| from slither import Slither | ||
| from slither.slithir.operations import * |
There was a problem hiding this comment.
There is a lot of things to import: https://github.com/crytic/slither/blob/dev-simil/slither/slithir/operations/__init__.py
😭
utils/similarity/encode.py
Outdated
| for contract in slither.contracts: | ||
|
|
||
| # Iterate over all the functions | ||
| for function in contract.functions: |
There was a problem hiding this comment.
Consider using contract.functions_not_inherited to simplify the upcoming #213
(function.contract == contract can then be removed)
There was a problem hiding this comment.
During training this could be useful to avoid repetition in the dataset, but I'm not sure this is what we want during testing. In that case, the user cannot specify a inherited function (since it will be missing). 😕
| else: | ||
| return _type.replace(" ","_") | ||
|
|
||
| def encode_ir(ir): |
There was a problem hiding this comment.
I would split this into two functions encode_ir and encode_value.
SlithIR is flat, so all the ir.value , ir.rvalue will never return an Operation
I removed that. It should be fixed now. |
- Fix import in utils.similarity.encore - Improve slither-similari help - Use contract.functions_not_inherited rather than contracts.functions + if function.contract == contract
This PR implements a new util,
slither-simil, to detect similar Solidity functions in a large amount of smart contracts using Machine Learning. More information and documentation is available here.