Feature extraction
Five texture properties across three distances and four angles produce 60 features per image.
Malware Image Classification
Finding patterns in malware images.
Explore the repository ↗Why this project?
Malware image representations can contain patterns that help distinguish families. This project explores classifying those images through texture features and supervised learning.
From input to output.
Images are converted to grayscale and resized to 64 × 64 pixels. Gray-Level Co-occurrence Matrix features describe their texture. Training scripts compare Random Forest, SVM, and XGBoost; the Flask interface uses Random Forest.
01Grayscale image
0260 GLCM features
03Random Forest
04Family + confidence
A simplified overview of the repository’s application flow.
Model comparison
Separate scripts support Random Forest, SVM, and XGBoost training and evaluation.
Prediction interface
An image upload interface shows the predicted family and confidence. Predictions below 60% confidence return an unclassifiable result.