Most security tools report what might be wrong. A Nullgap finding is different: it is only reported when it has been shown to work. This explainer walks through what that takes.
A lead is a claim
Nullgap starts the way an attacker would, by reading your code and your running app and listing what looks exploitable. Each of those is a lead. A lead is a claim, and claims are cheap. On its own it is not reported to you as a finding.
Four checks that all have to pass
A lead becomes a proven finding only when all four of these hold:
- The attack run works. The exploit is run and produces the harmful effect, such as one customer reading another customer’s invoice.
- The repeat works. A second run gives the same result, so it isn’t a fluke.
- A harmless control run doesn’t. The same request with an ordinary, harmless input behaves safely. This shows the effect comes from the attack, not from the app always behaving that way.
- An independent check accepts it. A separate review looks at the evidence and agrees it shows the flaw.
A status code alone never counts. A “200 OK” says a request was answered, not that anything harmful happened.
Why a sealed sandbox
Proofs run in a sealed sandbox built from your code, with no outside network. That keeps the attack contained, means nothing reaches your production systems, and makes the result reproducible. Live testing is the one exception by design: it reaches only the staging addresses on your scope list, read-only unless you allow state changes.
Anyone can replay it
Every proven finding ships with the exact requests and the sandbox output. Anyone on your team can replay it and see the same thing. You don’t have to take Nullgap’s word for it.
What “ruled out” means
When a lead fails the checks, it isn’t reported as a finding. It is shown as ruled out: tested and shown not to be exploitable. That list matters. It tells you what was covered, so a short findings list doesn’t leave you wondering whether anything was looked at.
Ruled out is not a guarantee that a weakness can never exist there. It means the attempts that were made did not work.
Then prove the fix
Nullgap writes a fix for each proven finding and re-tests it on a fresh checkout. Then the original exploit runs again against the fixed code. It has to fail before the finding moves to ✓ Proven fixed. Fixes arrive as pull requests; nothing is merged automatically.
Read more
- How proof works, with diagrams
- Testing access control across accounts
- Verified fixes