Contributor: Md Mahbub Islam Bhuiyan Macquarie University Department of Computer Science Email:[email protected]

This repository is a complete package for the experiment of Membership Inference VIA Backdooring In Image-Based Malware Classification for the degree of Master of Research at Macquarie University 2023.

1. In this experiment we have used dike dataset with binary files classified with Malware and Benign. You can access the dataset in this following link: https://github.com/Mahbub126/BackdoorMalware/tree/main/dike_dataset
2. BinaryToGrayConverter-Simple.ipynb program can convert any binary files of a directory to Grey Scale Images by providing the binary folder path as input. To convert a binary file to gray images you can use the script in this file: https://github.com/Mahbub126/BackdoorMalware/blob/main/BinaryToGrayConverter-Simple.ipynb
3. BackdoorImageGenerator with Black and white box.ipynb program can generate backdoor images in four different corners and trigger pattern can be white or black. To create backdoored image from gray image you can use the script in this file: https://github.com/Mahbub126/BackdoorMalware/blob/main/BackdoorImageGenerator%20with%20Black%20And%20White%20Box.ipynb
4. Train Backdoor Densenet.ipynb program trains Densenet ML model with clean samples or backdoor trigger samples. You can use the script in this file to train the Densenet ML Model: https://github.com/Mahbub126/BackdoorMalware/blob/main/Train%20Backdoor%20Densenet.ipynb
5. Train Backdoor Mobilenet.ipynb program trains Mobilenet ML model with clean samples or backdoor trigger samples. You can use the script in this file to train the MobileNet ML Model: https://github.com/Mahbub126/BackdoorMalware/blob/main/Train%20Backdoor%20MobileNet.ipynb
6. Train Backdoor VGG-16.ipynb program trains VGG-16 ML model with clean samples or backdoor trigger samples. You can use the script in this file to train the VGG-16 ML Model: https://github.com/Mahbub126/BackdoorMalware/blob/main/Train%20Backdoor%20VGG-16.ipynb
7. Train Backdoor VGG-19.ipynb program trains VGG-19 ML model with clean samples or backdoor trigger samples. You can use the script in this file to train the VGG-19 ML Model: https://github.com/Mahbub126/BackdoorMalware/blob/main/Train%20Backdoor%20VGG19.ipynb
8. Train Backdoor NasNetMobile.ipynb program trains NasNetMobile ML model with clean samples or backdoor trigger samples. You can use the script in this file to train the NasNetMobile ML Model: https://github.com/Mahbub126/BackdoorMalware/blob/main/Train%20Backdoor-NasNetMobile.ipynb
9. 30 Clean Test Sample images folder was used for testing clean and backdoored models. You can find the test samples in this folder: https://github.com/Mahbub126/BackdoorMalware/tree/main/30%20Clean%20Test%20Sample
10. Malware Images to add backdoor folder has the images which were backdoored and used in the training setup. You can find some the malware images in this folder to add backdoor trigger for training: https://github.com/Mahbub126/BackdoorMalware/tree/main/Malware%20Images%20to%20add%20backdoor%20for%20training
11. Testing_model_with_sample_images.ipynb program was used to test sample images for prediction. You can find the test script in this file: https://github.com/Mahbub126/BackdoorMalware/blob/main/Testing_model_with_sample_images.ipynb
12. To check the clean image trained model for testing you can find the trained models in this folder: https://github.com/Mahbub126/BackdoorMalware/tree/main/Clean%20Trained%20Model
13. To check the backdoored image trained model for testing you can find the backdoored trained models in this folder: https://github.com/Mahbub126/BackdoorMalware/tree/main/Backdoored%20Trained%20Model

© Md Mahbub Islam Bhuiyan