CRash Scope (CRScope) is a ML-guided crash classifier of a given JavaScript engine crash-dump file.

1. Install dependencies

   sudo apt update
   sudo apt install python-pip python-dev
   
   sudo pip install pandas==0.23.0
   sudo pip install scipy==1.0.0
   sudo pip install scikit-learn==0.20.3
   sudo pip install imbalanced-learn==0.3.3
   sudo pip install matplotlib==2.1.0
   sudo pip install python-docx==0.8.6
   sudo pip install XlsxWriter==1.0.2
   sudo pip install joblib==0.11
   

2. Clone CRScope

   git clone https://github.com/WSP-LAB/CRScope.git
   cd CRScope
   

1. Prepare the csv file that contains feature information of each crash instance.

   • We provide sample csv files in ./data directory.

2. Run CRScope

   usage: CRScope.py [-h]
                     [-e {chakra,v8,spider,chakra_v8,chakra_spider,v8_spider,all}]
                     [-s {under,over}] [-o OPTION] [-c {crscope,exniffer,combi}]
                     [-v VERSION]
                     datafile
   

   • -s, -o: Sampling method for balancing security and non-security data

     • -s over: Oversampling
       • -o 1: RandomOverSampler
       • -o 2: ADASYN
       • -o 3: SMOTE
     • -s under: Undersampling
       • -o 1: RandomUnderSampler
       • -o 2: TomekLinks
       • -o 3: CondensedNearestNeighbour
       • -o 4: OneSidedSelection
       • -o 5: EditedNearestNeighbours
       • -o 6: NeighbourhoodCleaningRule
     • reference: https://imbalanced-learn.readthedocs.io/en/stable/user_guide.html

   • -c: The choice for using which type of feature

     • crscope: CRScope
     • exniffer: Exniffer
     • combi: CRScope + Exniffer

   • -e, -v: For naming result directory and files

     • -e: A directory with the selected engine name will be created
     • -v: Result files with the version name will be created

   • For example,

      $ ./CRScope.py -e chakra -s over -o 1 -c crscope -v chakra_crscope_v1 ./data/chakra.csv
     

3. The final reports will be saved in result directory.

More details can be found in the paper.