README.md

Malware-Classification

This is the repo of our paper: Neural Network Innovations in Image-Based Malware Classification: A Comparative Study

This paper investigates the efficacy of deep learning models, particularly ConvNeXt V1 and V2, in the domain of image-based malware classification. The paper has been accepted at the AINA 2024 conference, Japan.

---

Steps to run the code

• 1. pip install -r /path/to/requirements.txt
• 2. run RQ1 notebook. (recommended on GPU)
• 3. run RQ2_RQ3 notebbok.

---

Code Files Structure

• RQ1 notebook:
  • Train and test different DNNs on two datasets: Malimg, and Malevis 224.
  • Store the results in a table and save it in a CSV file.
  • Visualize the results in a bar plot for effective comparison.
• RQ2_RQ3 notebook:
  • Train and test ConvNeXt V2 on six datasets.
  • Visualize the results in a bar plot for effective comparison.
  • Calculate the number of parameters for each pretrained model used in the existing works.

---

Data File Structure

• RQ1 checkpoints: contains the pretrained models used in RQ1.
• RQ1 checkpoints: contains the pretrained models used in RQ2.

---

Experimental Setup

---

Experimental Results

---

For any questions, please feel free to contact us 📧 📭:

• Hamzah Al-Qadasi: halqadas@asu.edu
• Djafer Benchadi: djafer@cvlab.cs.tsukuba.ac.jp
• Salim Chehida: salim.chehida@univ-grenoble-alpes.fr