"Releasing the Malware Analysis Dataset: Memory Forensics + ML for Ransomware Detection"

Aug 20, 2026 "AI Security"

I'm excited to announce the release of our Malware Analysis Dataset on Hugging Face — a comprehensive memory forensics dataset for ransomware research that combines the Volatility Framework, YARA rules, and machine learning for IoT/energy system cybersecurity.

This work was conducted with Sonika Malik and published at IEEE ICAIA 2026 (4th International Conference on Artificial Intelligence and Applications) under the title:

"Integrating Machine Learning and Memory Forensics for Enhancing Cybersecurity in IoT-Enabled Energy Systems"


What's in the Dataset

Component Scale
Memory Dumps 18 total (6 benign + 12 ransomware) × 4.5 GB each
Ransomware Families WannaCry, Cerber, GandCrab, Jigsaw, TeslaCrypt, ViraLock, BlueHowl, Annabelle, Vipasna
ELF Binaries 18 extracted binaries (benign + ransomware)
Volatility Scans malfind, pslist, psscan, pstree, psxview, vadinfo, vadwalk, yarascan, memmap
YARA Rules 100+ curated rules (packers, crypto, exploit kits, APTs, ransomware families)
VM Images Full Ubuntu 64-bit VM with snapshots
Code & Papers Automation scripts, Jupyter notebooks, IEEE ICAIA 2026 paper

Total: ~470 GB raw data + 33 GB analysis outputs — all public on Hugging Face.


The Research (IEEE ICAIA 2026)

Our paper "Integrating Machine Learning and Memory Forensics for Enhancing Cybersecurity in IoT-Enabled Energy Systems" presents:

  • Automated Volatility orchestration — plugin pipeline (malfind, vadinfo, yarascan, pslist) extracting behavioral features from raw memory
  • Feature engineering — process injection indicators, VAD anomalies, YARA match profiles, memory entropy
  • ML classification — benign vs. ransomware process identification using extracted forensic artifacts
  • 100+ YARA rules — custom signatures for ransomware family attribution (WannaCry, Cerber, GandCrab, etc.)
  • Real-world validation — tested on 18 memory dumps from controlled ransomware execution environments

This bridges the gap between traditional memory forensics (manual, expert-driven) and scalable ML detection for IoT/OT environments where memory analysis is critical but expertise is scarce.


Access Everything

  • Dataset (analysis): hf://datasets/Vedaang/malware_analysis — scripts, YARA, scan outputs
  • Raw data bucket: hf://buckets/Vedaang/malware-analysis-data/ — Benign_Analysis, Ransomware_Analysis, Scans, Virtual_Machines
  • Code bucket: hf://buckets/Vedaang/malware-code-base/ — automation, notebooks, papers
  • GitHub: github.com/Vedaang-Chopra/Malware_Analysis — documentation, quick start
# Quick access
hf download Vedaang/malware_analysis
hf sync hf://buckets/Vedaang/malware-analysis-data/Ransomware_Analysis ./ransomware

Why This Matters

IoT/energy systems are increasingly targeted by ransomware (Colonial Pipeline, etc.). Memory forensics is one of the few ways to detect fileless/in-memory attacks — but it's historically manual. This dataset + automation enables reproducible, ML-ready forensic research for the community.


Links


Co-authored with Sonika Malik. Shoutout to the open-source Volatility/YARA communities that make this work possible.