Sample article for layout review. It describes planned work and is not a published result.
Predictive malware intelligence depends on exposure to diverse behaviour. The corpus roadmap sets out a staged path to 100 million unique specimens, ingested in roughly equal annual stages of about 16.7 million each from 2027 to 2032.
Quality before quantity
Every specimen is hashed, deduplicated, and tagged with its source before it reaches training. Near-duplicate families are grouped so that evaluation splits do not leak closely related samples between training and test sets.
Sources
- Reputable public research repositories
- Synthetic samples generated in sandboxed container environments
- Governed community submissions through the THREATBYTES™ ingestion platform
Milestones are targets for measurement, not current claims. Progress will be reported against the roadmap as each ingest stage completes.