shuwang127 / patchrnn

The source code in the paper "PatchRNN: A Deep Learning-Based System for Security Patch Identification". This paper appears in the 2021 IEEE/AFCEA Military Communications Conference (MILCOM 2021), San Diego, USA, November 29–December 2, 2021.

SecurityPatchIdentificationRNN

Task: Security Patch Identification using RNN model.

Developer: Shu Wang

Date: 2020-08-08

Version: S2020.08.08-V4

Description: patch identification using both commit messages and normalized diff code.

File Structure:

|-- SecurityPatchIdentificationRNN
    |-- analysis                                # task analysis.
    |-- data                                    # data storage.
            |-- negatives                           # negative samples.
            |-- positives                           # positive samples.
            |-- security_patch                      # positive samples. (official)
    |-- temp                                    # temporary stored variables.
            |-- data.npy                            # raw data. (important)
            |-- props.npy                           # properties of diff code. (important)
            |-- msgs.npy                            # commit messages. (important)
            |-- ...                                 # other temporary files. (trivial)
    |-- SecurityPatchIdentificationRNN.ipynb    # main entrance. (Google Colaboratory)
    |-- SecurityPatchIdentificationRNN.py       # main entrance. (Local)

Dependencies:

pip install clang == 6.0.0.2
pip install torch == 1.2.0 torchvision == 0.4.0
pip install nltk  == 3.3

Usage:

python SecurityPatchIdentificationRNN.py