Machine Learning Research Assistant

The brief

The lab was classifying malware samples. Extracting features meant actually opening binaries in Ghidra and deciding what was worth measuring — which turned out to matter far more than the choice of classifier.

I piloted five-plus models and built the React dashboard the group used to look at the results.

What I did

• Piloted 5+ predictive machine learning models using decision trees and random forests for advanced malware threat analysis.
• Implemented a React analytics dashboard visualising 30,000+ malware datapoints, with features extracted through Python libraries and Ghidra reverse engineering.

Measured

Architecture

In hindsight

We compared models on accuracy alone for too long. On an imbalanced malware set that flatters a classifier that mostly guesses 'benign' — precision and recall told a much less comfortable story.

Stack