Malware family classifier from images

The brief

Malware analysts sort thousands of new samples into families. Using a public research dataset where malware binaries are already converted to grayscale images, train a CNN that recognises the family, and study which families it confuses. Good looks like: accuracy per family, a confusion matrix, and a short note on the method's limits against obfuscation.

What you will build

1. Explore the malware images by family
2. Handle family imbalance
3. Train a CNN and compare with a simpler model
4. Evaluate per family
5. Discuss limits such as packing and obfuscation

Free, open data to build it with

• Malimg malware images ↗ Kaggle

Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.