mahbub126/backdoormalware

How mahbub126/backdoormalware works

This repository is an experimental workflow for image-based malware classification and studying backdoors through membership inference. The documented path starts with malware binaries, converts them to grayscale images, optionally adds corner triggers, trains clean or backdoored image classifiers, and tests sample images with saved models. The README names five training architectures, preserved here as separate components. No notebook source was available, so relationships are limited to workflow statements in the README; implementation details and any distinct membership-inference analysis stage cannot be verified.

Main entry points: Clean Trained Model Clean Models · Backdoored Trained Model Backdoored Models · Testing_model_with_sample_images.ipynb Sample Image Testing · Train Backdoor Densenet.ipynb DenseNet Trainer

Components 17

Samples and Images

Dike Dataset dike_dataset
Binary Converter BinaryToGrayConverter-Simple.ipynb
Grayscale Images

Backdoor Preparation

Trigger Generator BackdoorImageGenerator with Black And White Box.ipynb
Training Malware Images Malware Images to add backdoor for training
Backdoor Training Samples

Model Training

DenseNet Trainer Train Backdoor Densenet.ipynb
MobileNet Trainer Train Backdoor MobileNet.ipynb
VGG-16 Trainer Train Backdoor VGG-16.ipynb
VGG-19 Trainer Train Backdoor VGG19.ipynb
NasNet Trainer Train Backdoor-NasNetMobile.ipynb

Testing and Models

Clean Models Clean Trained Model
Backdoored Models Backdoored Trained Model
Clean Test Samples 30 Clean Test Sample
Sample Image Testing Testing_model_with_sample_images.ipynb
Predictions

Other components

Researcher

Connections 26

Researcher → Dike Dataset: selects samples README.md
Dike Dataset → Binary Converter: converts binaries README.md
Binary Converter → Grayscale Images: produces images README.md
Grayscale Images → Trigger Generator: adds triggers README.md
Training Malware Images → Trigger Generator: provides malware images README.md
Trigger Generator → Backdoor Training Samples: generates triggered images README.md
Researcher → DenseNet Trainer: chooses training README.md
Researcher → MobileNet Trainer: chooses training README.md
Researcher → VGG-16 Trainer: chooses training README.md
Researcher → VGG-19 Trainer: chooses training README.md
Researcher → NasNet Trainer: chooses training README.md
DenseNet Trainer → Clean Models: trains clean models README.md
DenseNet Trainer → Backdoored Models: trains triggered models README.md
MobileNet Trainer → Clean Models: trains clean models README.md
MobileNet Trainer → Backdoored Models: trains triggered models README.md
VGG-16 Trainer → Clean Models: trains clean models README.md
VGG-16 Trainer → Backdoored Models: trains triggered models README.md
VGG-19 Trainer → Clean Models: trains clean models README.md
VGG-19 Trainer → Backdoored Models: trains triggered models README.md
NasNet Trainer → Clean Models: trains clean models README.md
NasNet Trainer → Backdoored Models: trains triggered models README.md
Clean Test Samples → Sample Image Testing: provides test images README.md
Clean Models → Sample Image Testing: tests clean models README.md
Backdoored Models → Sample Image Testing: tests backdoored models README.md
Researcher → Sample Image Testing: runs predictions README.md
Sample Image Testing → Predictions: returns predictions README.md

Drawn by GitDiagram on November 11, 2026 from the repository's file tree, README and source files. Read as Markdown ·