Mallory

Code for training and testing a CNN to learn bimap images of malware. The intuition is that a malware's binary can be turned into an image whose features can be learnt by a CNN. We use the MalIMG dataset to demonstrate the idea.

Data

• Download MalIMG dataset

• run Data/split.lua to split data into training, validation and test sets

• set datasetRoot in opt.lua

Training

> th main.lua -gpu -threads 4 -batchSize 30 

Testing

> th main.lua -testMode