Skip to content

Wapiti08/Defense_Evasion_with_GAN

Repository files navigation

Analysis-on-GAN

Authour Python GAN Ransomware License DOI


This is my project about Ransome Detection using GAN, the experiment was implemented by cuckoo.

Features:

  • read report files from specific location (set in cuckoo) and extract features from them
  • features engineering on choosing features with highest weights
  • generate matrix based on the features(occurences of features)
  • experiments with black attack for the GAN model
  • experiments based on different parameters, number of hidden layers, number of nodes
  • experiments based on ensemble classifiers
  • experiments based on ensemble neural networks
  • experiments on both ransomware dataset and generous malware

Pipelines:

- collect dataset:
    - please put the generated reports seperately under dataset folder
    - run the col_reports inside the processing (specify your own folder)

- generate raw features:
    - run the the feature_gen to generated seperate features

- get the total unique feature list:
    - run the feature_dict to get features list

- get the database based on feature occurrences
    - run the feature_fin to get dataset.csv

- classifier analysis (in classifiers):
   - analyse and filter the generated datasets and classifier algorithms

- model experiment:
    - test the models with filter features

For the ransomware and cuckoo configuration, please check the link cuckoo-download-instructions