Verified reachability
For the advisories that matter to your stack and configuration.
An advisory tells you a bug was fixed. It doesn't tell you whether a sibling path survived the patch — or whether your stack and config can reach it. The 0day hunter answers with a reproducer, or with the exact reason it doesn't apply.
AI can spit out vulnerability hypotheses all day. Somebody still has to decide which ones are real on your versions, your config and your apps. The hunter does that work — and won't call anything a finding until a command reproduces it.
Start from the fix: what changed, and which nearby paths it left alone.
Trace a candidate path through the source to the code your apps actually load.
Show the memory or logic fault under instrumentation.
Reach a fixed, benign command in an isolated lab. Nothing else counts.
Run it against the patched build and with the gadget removed. Both must fail.
Stamp the result with exactly one precondition tier.
Exactly one tier per result: STOCK, CONFIG-GATED or LAB-ONLY. Flip what the reproducer needs and watch the tier move.
What does the reproducer need?
A result moves up a tier only when a gate is removed — never by rewording it.
Our published PHP results: none are STOCK. We say so.
Reachable when the named gates hold. Check those gates against your stack.
For the advisories that matter to your stack and configuration.
Controls for each result, with its tier and stated limits.
So you know what you can safely deprioritize.
For the tools you already run: Sigma, Suricata, ModSecurity.
On patched PHP 8.4.27 and 8.5.12, a destructor-reentry path still reached freed memory in our local labs. We then followed it into Drupal, Joomla, Roundcube and WordPress and recorded exactly which conditions decide each one. None of it is a default-configuration exposure, and we say so. Our contribution is measurement: patched builds, a control matrix and detection, credited against the prior work we built on.
Hosting and platform providers, AppSec teams with large third-party surfaces, software vendors checking that their fixes are complete, and vulnerability-management teams that need verified exploitability, not just scores.
Our PHP research and its repository are already public, and the verification engine is planned as an open-source release. Teams subscribe to the verification feed: verdicts mapped to their inventory, VEX statements, detection packs and a replica of their configuration. Most of the feed is planned; talk to us about early access.
We're working with a small number of teams while we expand beyond PHP.