Note: you need to clone this repo using the --recursive flag since this repo has submodules, e.g., git clone git@github.com:ouerum/neural-FEBI.git --recursive

Neural-FEBI: Accurate Function Identification in Ethereum Virtual Machine Bytecode

1. The workflow of neural-FEBI

2. Prerequisites

• Build Solidity compiler from the source(https://docs.soliditylang.org/)
• Python 3.8

3. Running neural-FEBI

3.1 ground truth extractor

1. Use the project scripts/contractCrawler to download the Solidity code from EtherScan.
2. Compile the instrumented Solidity Compiler (instrumentedSolc/solc-0.4.25 and instrumentedSolc/solc-0.5.17).
3. Use the scripts/buildGroundTruth/compileContracts.py to compile Solidity Contract into EVM bytecode and annotation about how to construct function boundaries.
4. Use the scripts/buildGroundTruth/getGroundTruthBatch.py to extract the ground truth from the above annotation.

3.2 function entries identification

1. make sure the setting in FSI/utils/config.py are set.
2. Use the FSI/main.py to train the model to identify function entries of EVM bytecode.
3. Once the training is over, use the FSI/predict_operator.py to predict the function entries for the stripped EVM bytecode.

3.3 function boundaires detection

1. make sure the setting in FBD/fbdconfig.py are set.
2. Use the FBD/batch_analysis.py to detect function boundaries.

3. Dataset

URL: https://drive.google.com/drive/folders/1LsVEmKFALN9t2trWivPgEZI41lzeqEdu?usp=sharing

The dataset contains two parts :

• dataset/etherscan.zip: the solidity contracts crawled from Etherscan.
• dataset/ground-truth.zip: the ground-truth of function boundaries for given instrumented solc and contracts.

5. Publications

1. He, J., Li, S., Wang, X., Cheung, S.C., Zhao, G. and Yang, J., 2023. Neural-FEBI: Accurate function identification in Ethereum Virtual Machine bytecode. Journal of Systems and Software. https://doi.org/10.1016/j.jss.2023.111627