antiReaSUPER LAB

Super Lab / Benchmark edition 2026.11 / OWASP OAT-011 Scraping

Your site, analyzed. Then the fight, staged on its shape.

Paste a link to your site. The lab actually analyzes it - its listing structure, items per page, how many records it exposes, which fields each record leaks, what API surface shows - with a few read-only fetches, the way a scraper's first reconnaissance would. Then it stages the fight on a fair twin built to your site's shape: no defense vs antiRea, driven by attack and defense behavior measured in recorded benchmark runs. Your real site is never attacked. And every fail is published below, by design.

No link handy? Pick a standard site type instead:

Attack program (behavior from recorded runs)
Defenses to compare (multi-choice)

The analysis reads your page, its robots.txt and sitemap, and one item page - read-only, a handful of requests, nothing stored. The fight itself runs on a twin; on this web edition it is simulated from the benchmark edition's recorded runs, scaled to your site's real numbers. The downloadable Pro lab runs the live fight on your own machine.

Idle. The analysis and the arena wake up when a match starts.

How the frame stays honest

The analysis is real and it is reconnaissance-grade: the same handful of read-only fetches a scraper makes first. What it reports - structure, page size, record estimate - always names its basis (the site's own JSON-LD total, pagination math, a sitemap census, embedded page data, or a labeled lower bound). Estimates are labeled estimates.

Twins are basic market-standard sites with a correct basic setup, verified by a fairness probe on both arms; if a twin loses data, the cause is the attack method, never a careless victim. Battle outcomes on this page are simulations: the mechanics (budget stop, the key-gate wall, identity rotation, request counts, legit cost) come from the edition's recorded live runs, cited in every report, and a record counts as taken only under the same marker rule the live range uses. antiRea's known limits are published in its LIMITATIONS.md. The recorded 3.0.0 cells show both sides honestly: a per-identity budget alone loses to identity rotation, so 3.0 puts the tail behind server-issued keys - the listing path closes entirely, and what a key-holding harvester can still draw is capped by the issuance pool (measured at 10,000-record scale in the 3.0 validation run: 1,457 of 10,000 records).