All work
Illustration
Solution · Malware detection
For security teams and vendors · Since 2026
SHADOW
Malware detection from raw bytes
A foundation model that learns to recognise malware directly from the raw bytes of executables.
Status: Available to license or pilot
Shows the idea, not model output
The problem
Most malware detectors depend on features that someone chose by hand, which ties them to the families they were built around.
What was built
A foundation model trained directly on executable bytes, with no hand-built features.
What it does
- Learns from the raw bytes of executables, with no hand-built features
- Trained on a large, curated collection of malware and clean programs
- Aims to recognise malware families that were not in its training data
What we did
The model, its training data and the training runs.
- runs on your own samples
- Pilot
- hand-built features
- 0
- Windows executables
- PE