bin2ml: Turning Software Binaries into Machine Learning Ready Training Data

Josh Collyer

44CON 2025 · Day 1 · Main Track

Machine learning applied to binary analysis is a field drowning in interesting ideas and starved for good training data. Papers proposing neural network approaches to tasks like function similarity se

AI review

Collyer built a tool the field actually needed, makes the right argument about why ML-for-binary-analysis research is broken, but this is a tooling talk at a security conference rather than a research talk — the science lives elsewhere.

Watch on YouTube