Julia malware classifier

XGBoost on EMBER 2018 in Julia: 88.17% accuracy, 0.9581 AUC from byte histograms.

End-to-end malware classification pipeline in Julia: ingest, EDA, normalization, gradient-boosted trees, and publication-quality metrics. Benchmarked on the 1M-sample EMBER 2018 set using a minimal 256-d byte histogram baseline to prove Julia’s viability for high-performance security data science.

Highlights

• 88.17% accuracy / 0.9581 test AUC on a minimal feature baseline
• Shannon entropy utilities exposing malicious vs benign structure deltas
• Reproducible Dockerized training over 600k skewed binary samples