Run a website vulnerability scanner against a real application and you'll get a number back. Four hundred issues. Maybe four thousand. It feels like progress — look how much it found! But that number is not a measure of security work done. It's a measure of security work created, and handed to you.
Because now someone has to go through all four hundred. Figure out which ones are real. Which ones are reachable. Which ones actually matter. That triage is the expensive part, and the scanner just made it your problem.
Recall is easy. Precision is the hard part.
Scanners are tuned for recall — catch everything that might be an issue, because missing a real bug is the cardinal sin. That instinct is reasonable in isolation. The problem is what it does to precision: with no step that confirms exploitability, the tool can't distinguish a live, reachable flaw from a theoretical one in a code path nothing ever calls. So it reports both. And both land in your queue looking identical.
A finding you have to validate yourself isn't a finding. It's a lead. The work of turning a lead into a finding is exactly the work you were trying to automate.
The math of false positives
Say a scan returns 400 issues and, generously, 10% are real and exploitable. That's 40 real bugs buried in 360 false or unreachable ones. To get to the 40, a human reads all 400. At even five minutes of triage each, that's over 33 hours of work — most of it spent confirming that something isn't a problem.
Two things go wrong from there. First, the triage cost is so high it often doesn't happen — the report gets skimmed, or shelved. Second, when 90% of what a tool says is wrong, people stop trusting the 10% that's right. The real bugs get the same shrug as the noise, and ship anyway. Noise doesn't just waste time; it trains people to ignore true positives.
What “validated” actually means
A validated finding is one the tool has already proven before showing it to you. Concretely:
- It's reachable. The vulnerable path can be hit from where the tool says it can — not just present somewhere in a dependency.
- It's exploitable. There's a request, or sequence of requests, that actually triggers the behavior.
- It has demonstrated impact. The tool can point at what the exploit yields — data returned, access gained, a boundary crossed.
- It comes with a reproduction. If it was confirmed, the exact steps already exist. They ship with the finding.
When those four hold, triage collapses. You're not asked “is this real?” four hundred times. You're handed a short list where the answer is already yes, and asked the only question that needs a human: what do we do about it?
This is a different metric
The industry trained itself to measure tools by volume — findings per scan, CVEs in the database, coverage of signatures. Those are vanity numbers. The metric that predicts whether a tool saves you time is the fraction of its output that's true. A tool that surfaces 15 real bugs with a 95% hit rate is worth more than one that surfaces 400 issues at 10%, even though the second “found more.”
This is why validation, not discovery, is the thing worth paying for. Discovery is close to a commodity — lots of tools can flag a maybe. Confirming the maybe, cheaply and at scale, is the hard problem, and it's the one that changes the economics of vulnerability management.
How HUNTR approaches it
HUNTR treats validation as a required step, not an optional one. It maps the attack surface and tests it class by class like any thorough tool — but a potential issue isn't a finding until the engine has confirmed it's real and drafted the reproduction. False positives are dropped before they ever reach you. What lands is a short list of validated, report-ready vulnerabilities, each with repro and impact. You review and submit; it never submits for you.
We'd rather hand you ten bugs you can trust than four hundred you can't. The benchmark explains how we score that — recall andfalse-positive rate, against targets where the answers are known — and the platform page walks the pipeline end to end.
Frequently asked
- What is the difference between a vulnerability scanner and validated findings?
- A vulnerability scanner reports anything that matches a known signature or looks suspicious, without confirming it is exploitable. Validated findings are issues a tool has actually proven — with a working reproduction and demonstrated impact — before showing them to you. The first is a to-do list of maybes; the second is a short list of real bugs.
- Why do vulnerability scanners produce so many false positives?
- Scanners optimize for recall: flag everything that might be an issue so nothing is missed. Without a validation step they cannot tell a reachable, exploitable flaw from a dormant one, so they err toward reporting. The cost of that choice lands on whoever triages the output.
- How do you measure the real cost of scanner noise?
- Count the human hours spent triaging findings that turn out to be false or unreachable, plus the real bugs that get buried in the pile and shipped anyway. A tool with a 90% false-positive rate isn't saving you time — it's generating a second job: sorting its output.
- Do you still need a scanner if you have a tool that validates?
- Scanners remain useful for asset inventory, compliance evidence, and patch tracking. For actually finding and confirming exploitable vulnerabilities, a validating approach is what removes the triage tax. They answer different questions.