This is an attempted fixed version of SySeVR: https://github.com/SySeVR/SySeVR. There are still some known issues.

• Please follow the instructions in config.sh. Don't run it directly as it contains some branching instructions. Read and follow it.
• For a quick/sanity testing of the environment, you can remove most of the folders after executing ./split-sard.sh, which will dramatically reduce the data processing time.
• You need to manually change the flag in bgru.py to switch the model between training and testing. Currently it's set to training. See corresponding switches in line 69 and line 110.

• The data labeling script may end up assigning every slice the same label. For a quick fix, please refer to a new simplified labeling script provided in the src_vf folder.

The VF extension simplified the data labeling process a bit, and provide a basic graph neural network pipeline that performs classification. Note that this is NOT a re-implementation of the original SySeVR.

The processed data (for SARD) can be downloaded here: Google Drive. Use the notebook in src_vf to continue with learning or manipulate the data processing.

The latest processed data (that is provided to the network, with pre-processing like new tokenization etc., as of July 27, 2021) can be found here: Google Drive.

• SySeVR_fix: https://github.com/kingnop/SySeVR_fix
• environment setup: https://github.com/tophertimzen/vagrantized-Joern/blob/master/setup.sh
• folder splitting:https://stackoverflow.com/a/36511395/14471117
• SySeVR guide: https://blog.csdn.net/weixin_42072280/article/details/106658014