misraya / concept-drift-robustness

Analyzing robustness of ML algorithms to concept drift on Windows PE and Android Malware datasets

concept-drift-robustness

Windows PE datasets:

First create a folder named data if it does not exist:

mkdir data

Then, download the data from the following links:

BODMAS

https://drive.google.com/drive/folders/1Uf-LebLWyi9eCv97iBal7kL1NgiGEsv_

EMBER

A version of EMBER preprocessed by BODMAS authors can be found here, only ember related files are needed:

https://drive.google.com/drive/folders/12DMPeh8DA2ukPATnHX4K__shWFJIiBN5

Place the .npz data and .csv metadata files in the data folder.

Android Malware datasets:

KRONODROID

Download the zip files from the following link, and unzip them in the data folder:

https://github.com/aleguma/kronodroid/tree/main/emulator

---

Training

Training script can be used to train a specified model on a selected dataset, with optional arguments such as train_start_date, train_end_data, and more. For example, to train a decision tree model on the bodmas dataset with default arguments, use the following command:

python src/train.py --model_type DT --dataset bodmas 

Replicating results:

First download the data from the above instructions, then run the following

python src/run_all_experiments.py
python src/kronodroid_experiments.py