#035Technical4 min
Tell a confirmed vulnerability from a possible false positive
Read evidence and confidence, investigate contradictions, and report an incorrect detection with a useful explanation.
Hackaiz team · updated 21/11/2026

01Separate severity from proof
A false positive is an incorrect detection, not simply a vulnerability you consider low priority. Start by separating two questions: how serious would the reported weakness be, and how well is its presence supported? Hackaiz shows severity as Danger, To fix or To watch. It also displays High confidence, Medium confidence or Low confidence. A lower confidence finding is not automatically wrong, and a severe label does not remove the need to inspect its evidence.
The evidence badge provides another reading aid. Proven appears when the finding has evidence and High confidence. Evidence available means supporting content exists without that combination; Not conclusive means there is no evidence text for that badge. These are product indicators to help review a result, not independent certification or a guarantee that a particular exploit works in every deployment. A finding supported in source code is not necessarily a demonstrated incident on the live service.
02Check the exact resource and version
Open Projects, choose the project and open the finding. Compare What was found with Evidence, the location and the analyzed version. For a repository finding, inspect the indicated code in its real execution context. For a website finding, examine the tested address and observed response. Do not reject a result just because there is no source line: an externally observed behavior can be documented by an address and request instead.
A plausible false-positive explanation must contradict the reported weakness. For example, a detected value may be a deliberately public identifier rather than a privileged secret. Support that distinction with its intended role and actual permissions, without sharing a live secret. By contrast, saying that users cannot see a button does not refute a finding that the server accepts an unauthorized request. An interface restriction and a server-side access check protect different boundaries.
- 1Confirm that you are reviewing the same resource and analyzed version.
- 2Inspect the raw observation before relying on its summary.
- 3Look for a concrete contradiction, such as a non-secret public identifier or code that is not executed.
- 4If the evidence remains ambiguous, investigate rather than dismissing it solely to clear the active list.
03Use available verification without expanding scope
When the finding contains suitable evidence, the detail page may show Replay the proof, with The flaw exists if and It is fixed if, or a test that you can copy into your repository. These aids are conditional; not every finding includes a replay command or generated test. Review the target and expected result before running anything. A test aimed at your test server should not be pointed at unrelated production services.
All website and API analysis requires domain-control proof through the _hackaiz DNS TXT record, the HTTPS file /.well-known/hackaiz-verification.txt or having the project’s AI create the file. Keep any follow-up within your authorized scope. Additional analyses consume credits. Unticked checks are not executed or billed, so their absence from a later report cannot settle a dispute about an earlier finding. Compare like-for-like observations instead of assuming that any empty result disproves the original.
04Report a technical error, with a reason
If the result is incorrect, select Report a false positive on the finding detail. The form asks Why is this detection wrong? Describe the mismatch between evidence and reality, then use Send report. A useful explanation identifies the resource, the behavior the rule inferred and the fact that contradicts it. Avoid passwords, tokens, complete personal records or other sensitive material in the explanation. Concise technical reasoning is more useful than a general statement that the application is safe.
The report marks the finding as disputed rather than deleting its history. The interface shows Reported as a false positive and offers Undo report if your conclusion changes. Reporting does not apply a security fix or prove the whole application safe. If you discover that the weakness is real, restore the finding and address it. Where GitHub remediation is available, changes are draft pull requests reviewed and approved by a human, never merged automatically.
Frequently asked questions
Take the next step
Was this article helpful?