C SOFTWARE SECURITY
A patch says the bug is fixed. We check that it is.
01 — THE RISK
Scanners find the hole and a developer writes the fix. Nobody shows that the fix actually closes it.
02 — WHAT IRONPROOF DOES
Ironproof writes a test that triggers the bug, runs it before and after the fix, and tells you whether the fix closes it, within stated bounds. Each result ships as a dossier you can re-run.
03 — WHAT YOU RECEIVE
Three verdicts. Each one ships with its verification dossier.
Never a green checkmark. Every result is one of three, and each comes with what was used to reach it.
Proven.
The property holds within the stated model and its assumptions.
You get the obligation — SMT-LIB for policy proofs, the harness and its bounds for C — to re-run yourself.
Refuted.
A concrete counterexample shows exactly how it breaks.
You get the counterexample, to replay yourself.
Not demonstrated.
We could not conclude. Said as such, never shown as a pass.
You get the reason it stopped, recorded with the result.
04 — TODAY
20 real fixes drawn at random from projects such as curl and libssh2: 13 closed, 3 where our test was fooled, 4 not analyzed (October 12, 2026).
Others sell trust.
We sell the right not to trust us.
Bring one case. Leave with the evidence.
Tell us what you need checked. We answer with what we can prove, and what we cannot.
ALL THREE APPLICATIONS