This project provides the coresponding code to the paper "Resisting Adversarial Examples using Hashing-based Deep Neural Networks in Malware Detection"
- Python 2.7
- Numpy 1.11.3
- Matplotlib 2.0.0
- Tensorflow 1.3 or 1.4
- Scikit-Learn 1.0.0
- Jupyter notebook
In root direcotory, we do joyful examples on synthesis dataset:
- generate adversarial examples:
gen-adv-smps.ipynb - construct DNN graphs
graphs.py - Multi-index hashing based DNNs
InH.ipynb - Local forest hashing based DNNs
LFH.ipynb - Joint index hashing and Denoising auto-encoder
JID.ipynb - Joint locah forest hashing and Denoising auto-encoder
JFD.ipynb - other files
utils.py learning_hashing_by_RF.py
We recommend to run gen-adv-smps.ipynb
first to obain adversarial examples, and then perform any one of InH.ipynb
, LFH.ipynb
, JID.ipynb
and JFD.ipynb
In drebin directory:
There are source codes for a series expriments on drebin. We need apply the DREBIN dataset at here: drebin.
In pdfrate directory:
We provide the robust support vector machine method.
We conduct the experiments on the CPU server which 64 cores CPU (2.4GHz) and shared 150G RAM.