Skip to main content
The Panaseer logo shows a white square and a yellow square around the initial P. To the right of the P there is the copy written ‘anaseer’.

"How exposed are we?" PHP2Shell demonstrates why a vulnerability count isn't enough

"Where do we have exposure? As opposed to where do we have a vulnerability that needs to be patched?"

Lucy Caiger
read

That's how one security leader described their team's work during PHP2Shell.

When PHP2Shell was disclosed on 21 October, security leaders got the same question from their boards: how exposed are we? The quickest answer (and the one most teams default to) is a count of affected devices, but that count only tells part of the story.

We tracked affected devices daily across a cross-section of our enterprise customers, and spoke to the security operators working through the response. Teams initially patched fast, fixing 21% of affected devices within 24 hours, but then progress slowed. Open affected devices peaked on Day 4 and, by Day 13, were only 12% below that peak.

Alongside mobilizing teams to patch, a lot of time and effort can initially go into knowing what you're dealing with:

  1. Where is the vulnerable software, and can your scans see it?
  2. How critical are the affected devices to the business?
  3. How likely is each affected device to be exploited?
  4. Can the affected devices be patched? If not, what compensating controls are in place?

For each question, we share what the PHP2Shell data and the operators we spoke to revealed, and where to focus your effort now, so you're ready before the next one lands.

A quick recap: PHP2Shell refers to two critical, unauthenticated remote code execution flaws in PHP (CVE-2026-109618 and CVE-2026-109619), fixed on 21 October after details leaked a day early. CISA added both to its Known Exploited Vulnerabilities catalog five days later. Where scored, affected detections in our data carried an Exploit Prediction Scoring System (EPSS) probability close to the top of the scale.

1. Where is the vulnerable software, and can your scans see it?

Both stats point to gaps in what your scanners see, and when they last saw it. Those gaps cause three key problems:

  • Blind spots. A device that has never been scanned won't show up, however vulnerable it is.
  • A lag in understanding. When a critical CVE lands, devices scanned outside SLA can take days to show whether they're affected.
  • Out-of-date fix status. Until a device is rescanned, you can't confirm a patch has worked, so you look more vulnerable than you are.

All three leave your answer to the board resting on missing or out-of-date data.

One way to find coverage gaps is to compare scanner results with other device-centric sources, such as the CMDB, EDR and configuration management tools, and look for devices those sources can see but the scanner has never reached.

While you wait for scans to catch up, a software inventory drawn from EDR, asset management or configuration management tools can show where the affected software is installed, and which version, within hours of disclosure. Cross-check it against scan results as they arrive.

One team's scans came back suspiciously clean. They found that their web application firewall (WAF) was blocking the scanner from reaching the hosts behind it. Scans were happening, but the vulnerabilities weren't visible. Check that scans reach their targets and complete: scan from behind the WAF or use agents, and use authenticated scans where you can, so installed software is visible.

Affected products were also added after disclosure. Firewalls, VPN gateways and load balancers from several major vendors bundle their own PHP, so each vendor confirmed impact and issued its own patch over the following days. Scanners could only flag them once vendor-specific checks existed. Vendor advisories were still being updated a week after disclosure, so even a Day 7 count can miss them. Keep tracking after the first wave.

Start here:

  • Compare your full device inventory, from every source you have, against scan records.

  • While scans catch up, use a software inventory to find where the affected software is installed, then cross-check against scan results.

  • Track three numbers: devices never scanned, devices scanned outside SLA, and scans that failed or ran unauthenticated.

  • Treat suspiciously clean results as a reason to dig, not to relax.

2. How critical are the affected devices to the business?

Business process, service and location, usually from the CMDB, drive how security teams prioritize and what they report upward. Without that link, a vulnerable server behind a payment service looks no more urgent than a test box nobody uses. Everything starts to look equally urgent, which is how exposure-whelm* takes hold.

Start here:

  • Measure what share of your assets have no business-service tag, and close that gap for your critical services first.
  • For each business service, map the applications it depends on and the infrastructure underneath, so any vulnerable host can be traced to what it puts at risk.

*More exposures appearing, and being exploited, faster than organizations can realistically prioritize and fix them.

3. How likely is each affected device to be exploited?

Start with reachability: does the device face the internet, and is it still active? CERT-EU and the UK's NCSC advised isolating internet-facing systems, or taking them offline, where patching had to wait. That advice assumes you know which systems face the internet. About 70% of affected devices we tracked had no internet-facing or internal classification at all.

PHP's advisory focuses on installations that serve web requests, so a developer's laptop isn't the same risk as a public web server running mod_php. Operators used the file path where PHP was found to tell them apart.

Start here:

  • Tag every asset as internet-facing or internal, starting with your critical services.
  • When a critical CVE lands, read the advisory for what makes it exploitable (for PHP2Shell, PHP serving web requests) and take that into account.

4. Can the affected devices be patched? If not, what compensating controls are in place?

Initial fixes ran through the remediation and ticketing routes teams already had. After Day 9, though, fixes had stalled.

What was left were the hardest platforms. Devices on end-of-life operating systems were fixed at 12% by Day 13, against 43% for everything else: less than a third of the rate.

End-of-life systems often can't be patched at all. For them, the answer may not be a patch, but rather a compensating control, backed by a formal risk decision. A compensating control might segment the application, or cut it off from the internet until it can be fixed.

Some protection may already be in place before you patch. In at least one case, existing WAF signatures flagged PHP2Shell-style requests before any dedicated signature existed. WAF coverage reduces exposure, but you can only rely on it if you know which assets sit behind it.

Choosing between a fix and a compensating control is ultimately a cost decision, but that decision needs evidence that the control is in place and mitigates the risk.

Start here:

  • Agree in advance how end-of-life and hard-to-patch platforms will be handled.
  • Define the process: who chooses between a fix and a compensating control, what evidence they need, and when exceptions expire.
  • Fix what puts critical services at risk first, and make sure what can't be fixed yet is understood and covered by compensating controls.

Before the next critical CVE

PHP2Shell won't be the last critical vulnerability to arrive with public exploit code, and next time your answer will depend on what you already know. Get these four areas right and you can prioritize remediation where it matters, and give your board an answer you can defend: what's reachable and still open, who owns it, and what you don't need to worry about.

About the author

Lucy Caiger

Lucy is Director of Customer Success at Panaseer and leads the Panaseer operators forum.