Detection quality, measured in the open
A detection company that does not publish its accuracy is asking you to take its word for it. We would rather show the numbers and the method behind them. The results below come from a benchmark our team can re-run, measured on fresh threat-feed URLs that are still serving a real page at scan time (a content-aware liveness check, not just an HTTP 200), against a benign control set. Every figure is regenerable, and we publish a result only after human review.
Measurement in progress
We publish a detection-quality result here only after human review, and only when it represents fresh, live ground truth. Many threat-feed URLs are already dead or cleaned by scan time, so a naive number would understate real quality. Until an approved, representative run is ready, no accuracy figure is shown here.
What we measure
Each automated run scans freshly reported threat-feed URLs that a content-aware liveness check confirms are still up at scan time, plus a benign control set. The headline number is precision and recall on the phishing family, the job this scanner is built for, never a blended figure across unrelated threat types.
Exactly what is pending
A representative benchmark run has not been approved yet.
When this page updates
Benchmark runs execute automatically about every 24 hours. This page reads the live result feed, so an approved run shows up here immediately, with no site release in between.
Known limitations
- Threat-feed URLs decay fast; only samples confirmed live at scan time count.
- Results describe the URL and phishing scanner, not the contract or repository scanners.
- A single run is a snapshot. Runs repeat on the cadence above and replace older candidates.