A technical publication surface for detection science, reproducibility, benchmarks and adversarial behavior. Every publication carries an explicit evidence class so readers can distinguish validated project evidence from synthetic, experimental and planned work.
5 publications
The flagship ThreatFade research protocol for reproducible evaluation of fade scenarios, benign transient fades, provenance and limitations.
A pinned benchmark contract covering datasets, splits, metrics, provenance, performance measurement and publication rules without inventing results.
The ThreatFade detection thesis: reduced observability can be modeled as a behavioral deviation rather than assumed benign.
How rolling Shannon entropy and statistical deviation participate in ThreatFade's evidence-first pipeline.
Why deterministic benchmarks, project validation and independent assurance must remain separate evidence classes.
Repository validation, synthetic fixtures, independent evaluation and planned research remain separate evidence classes. Passing tests is not represented as universal detection accuracy.
The first Phase 16 study pins the dataset, engine commit, configuration and metrics before any results are published.
Read protocol →A public, versioned challenge for reproducible fade-detection experiments. The leaderboard stays empty until submissions are actually evaluated.
View challenge →