NSA outage blamed on PHP2Shell patching
Officials say patching for PHP2Shell knocked the NSA's classified network offline for more than three days. What's confirmed and how to check your own PHP.

On this page 4 sections ▾
TL;DR
- Multiple officials now back the DNI's PHP-patching explanation, though the NSA's own statement still only says routine maintenance and nobody has said which systems failed.
- If you patched PHP to 8.4.27 or 8.5.12, those builds have their own crash bug, fixed on November 19 in 8.4.28 and 8.5.13: update, don't roll back.
In October, the NSA joined the FBI and CISA in telling everyone running PHP to patch PHP2Shell right away or take their servers offline. Then, on November 17, NSANet, the agency's own classified network, went down, and the Director of National Intelligence is blaming the patching for the NSA outage.
Quick facts
Amazon saw attackers going after PHP2Shell four days before PHP shipped the fix.
The NSA told staff they could come back to work as normal, then shut its campus to all but essential personnel a few hours later.
#What happened to NSANet
NSANet is the network NSA staff use to share classified information inside the agency, and six people familiar with the matter told CNN it was down that Tuesday. Here's how the day went:
- A voicemail for employees declares a "code yellow," the second-highest of three alert levels at the Fort Meade, Maryland headquarters: non-emergency staff arrive in staggered shifts, emergency staff as usual.
- The NSA blames "scheduled routine maintenance" and says its systems have multiple layers of backup.
- The NSA closes its campus to all but essential personnel while it works to restore the network.
On Wednesday, non-essential day-shift staff in Maryland were told to stay home unless a supervisor called them in. Four sources told CNN the outage also hit NSA Georgia at Fort Gordon, which handles signals intelligence and cyber operations for the Middle East, where the US is at war with Iran.
Two sources told CBS that NSANet was back up by Wednesday, but it wasn't over. CNN later reported that large numbers of employees were locked out for more than three days.
It was resolved by Friday, November 20. That morning's employee voicemail said NSA Washington was "operating under normal conditions with exceptions," and staff came back in staggered shifts to avoid overloading the system with too many logins at once, with most expected back by midday. A senior administration official acknowledged that the patches needed to end the outage took longer than anticipated.
#Why officials blame a PHP patch
Director of National Intelligence Jay Clayton was the first official to name a cause, and he's still the only one to do it on the record. Here's his full answer on Fox Business:
If you go back about 10 days ago, we had what was labeled as PHP, an update to a web interface system. And the patching that has gone on following that has caused some of these systems to not come back up when you patch them as expected. That is what has happened at the NSA today.
PHP2Shell is a pair of flaws in PHP itself that let attackers run code on a web server without logging in, which we covered in our PHP2Shell post. PHP shipped the fixes on October 21 in 8.5.12, 8.4.27, 8.3.36 and 8.2.35. Five days later, CISA added both flaws to its Known Exploited Vulnerabilities list and gave federal civilian agencies three days to patch.
What we still don't know
Multiple officials briefed on the outage have since told CNN it appeared connected to how the NSA handled its PHP updates, but they spoke anonymously, and the agency's own statement still only says "scheduled routine maintenance." Nobody has said which systems failed or why they wouldn't restart. Clayton's dates don't line up: he said "about 10 days ago," but PHP2Shell's fixes came out on October 21, almost four weeks earlier. And a U.S. intelligence official told CBS that the volume and speed of changes can overwhelm aging infrastructure built for an earlier internet, so the problem may be bigger than one patch.
#What the NSA outage says about emergency patching
Patching fast was still the right call, since attackers were using PHP2Shell before the fix existed. What the NSA's week shows is how patching itself can turn into an outage:
- Check it came back, not just installedA package manager saying the update worked doesn't tell you the service restarted, and services not coming back is exactly what Clayton described. Test that each one answers before you patch the next batch.
- Patch in wavesThe intelligence official pointed to the volume and speed of changes. Patching one group of servers at a time means a bad result only takes down that group.
- Bring users back slowlyExpect a login rush when a system comes back. That's why the NSA staggered arrivals, and it's worth planning for before you restore anything.
These lessons apply right away: on November 19, PHP shipped 8.5.13, 8.4.28, 8.3.37 and 8.2.36, another security release with seven critical fixes, a month after PHP2Shell.
#Did your own PHP2Shell patch break anything?
Nothing ties the NSA's trouble to a specific bug, but the PHP2Shell fixes for two branches did ship with one. On PHP 8.4.27 and 8.5.12, the garbage collector can crash PHP when an app uses SimpleXML. PHP's developers traced it to a separate change in those releases, not to the security fix itself.
Two commands tell you whether the crash bug applies to you: which PHP build you're on, and whether SimpleXML is loaded. Here's the output from an Ubuntu 24.04 test machine:
student@opnf:~$ php -v PHP 8.3.6 (cli) (built: Oct 19 2026 15:16:58) (NTS) Copyright (c) The PHP Group Zend Engine v4.3.6, Copyright (c) Zend Technologies with Zend OPcache v8.3.6, Copyright (c), by Zend Technologies student@opnf:~$ php -m | grep -i simplexml SimpleXML
If yours shows 8.4.27 or 8.5.12 and lists SimpleXML, you're on a build with the crash bug. This machine's 8.3.6 isn't one of the builds named in the bug report. An old-looking number like that doesn't mean it's missing the PHP2Shell fix: our PHP2Shell post shows how to check that from the package changelog.
If your PHP workers are crashing
PHP fixed the bug in 8.4.28 and 8.5.13 on November 19. These steps, from the WordPress Hosting Handbook's note on the bug, cover spotting it and getting onto the fixed release:
- Look for crashed workers in the PHP-FPM logThe crash shows up as a child that "exited on signal 6 (SIGABRT)". The clearer "zend_mm_heap corrupted" message only appears if the pool sets
catch_workers_output = yes. - Don't downgrade to stop the crashes8.4.26, 8.5.11 and earlier are still open to PHP2Shell, which attackers are already using. Occasional crashes are the safer trade, even if you can't tell whether this bug is the cause.
- Update to 8.4.28 or 8.5.13Install them from php.net, or your distro's update once it lands, then restart PHP-FPM so every worker loads the new build.
Patching a fleet, checking that services actually came back and reading worker logs are everyday server administration. If that's new to you, the Security Pro ProCourse teaches Linux server administration in a simulated terminal in your browser, and Level 1 is free.
Was the NSA outage a cyberattack?
What is NSANet?
Is the NSA outage connected to the other outages this month?
What this means for you
If you patched PHP for PHP2Shell, check that every service using it actually came back and that PHP-FPM workers aren't crashing on 8.4.27 or 8.5.12. Move to 8.4.28 or 8.5.13, which fix the crash, and never downgrade to a vulnerable build to stop it.
SourcesCBS NewsCNNCNN, November 18TenableWordPress Hosting HandbookPHP changelogPrecursor SecurityCNN, November 20
Get opnF in your inbox
Tech news for IT professionals that actually teaches you something.
Our promise to you: we'll be conscious about email quality/frequency, and your address will never be shared/sold.