One of the early targets of PHP2Shell exploitation was the firewall. On October 28, Palo Alto Networks confirmed that attackers were exploiting the two PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619, against PAN-OS management interfaces exposed to the internet. Both flaws let anyone execute code on the web server without authentication. The disclosure itself went a little wrong. Details leaked on October 20, ahead of the coordinated release, and the PHP project shipped an emergency patch the next day.
The vulnerabilities require no authentication, but they are not easy to exploit reliably. It takes skill and additional knowledge about the target system. Where it matters a lot is appliances, which run identical, often outdated software that an attacker can study in advance. That is the bigger picture. An awful lot of web applications control critical devices in your network. Almost anything is a web application to some extent. That often goes unrecognized, and as a result those applications aren’t adequately protected.
Everything is a web application. That is the motto of the SANS course I teach, SEC522: Application Security: Securing Web Applications, APIs, and Microservices. Web application security has traditionally been treated as a developer issue, yet anyone protecting a network must understand how web applications work to properly defend them.
Why Appliances Are the Softer Target
On ordinary web servers, address space layout randomization (ASLR) is one of the features that makes exploitation harder. It shuffles where code and data sit in memory, so an exploit can’t rely on fixed addresses. VulnCheck’s early analysis found the bugs exploitable on ARM64 systems and on x86 systems without ASLR. Few public exploits bypassed ASLR on x86-64, and the one VulnCheck confirmed worked only against a single PHP build. VulnCheck’s own exploit chained a memory leak to get code execution with ASLR enabled.
Appliances change that math, not because the protections are missing, but because every device on a given software version runs the same binaries, so an attacker can work out memory offsets in advance. The PAN-OS exploit still had to defeat ASLR, leaking the libphp.so base address through the Set-Cookie header, and SELinux confined the resulting shell. PAN-OS also ships end-of-life PHP versions (8.0.30 and 7.4.33). Israeli security firm Dream, working with Palo Alto Networks on the investigation, found the attacker had obtained PAN-OS firmware, decrypted it, extracted the PHP libraries, and built precise memory profiles for nine software versions. The actor then chained PHP2Shell with a Linux kernel flaw (Fragnesia, CVE-2026-46300) to gain root on 25 systems (24 firewalls and one management server) in the U.S., Japan, Taiwan, and South Korea.
Web application security tends to focus on large, cloud-based applications. It’s often the little things, the mosquitoes on the internet, that do you in. Appliances run predictable, often outdated software, and they are everywhere.
HTTP Is More Complex Than Your Defenses Assume
HTTP is a very complex protocol with lots of options. To upload a file, a client doesn’t have to send it in one piece. It can send the body in chunks without declaring its total length, and, if the server decompresses request bodies, compress it, which makes it difficult for a web server to calculate how much memory it needs to store it. Memory-safety bugs remain common as a result, in web applications and in the web servers running them. In PHP2Shell, a crafted multipart upload pushes PHP past its memory limit while it is still parsing the request, before any application code runs. The bugs are in what PHP leaves behind when it aborts the request: uninitialized request memory (CVE-2026-109618) and a pointer to freed memory (CVE-2026-109619). The request doesn’t need to be large, either. Deeply nested, empty file fields make PHP allocate far more memory than the body contains, so a request well under the default 8 MB post_max_size can still exhaust the default 128 MB memory limit.
Defenders must understand the protocol to know what these vulnerabilities look like, for example to write signatures that detect the exploits. The same complexity limits the middle boxes we rely on. They can even interfere in ways nobody planned. Intruder found that nginx in front of PAN-OS buffered chunked request bodies, which broke public exploit chains. Web application firewalls are often not adequate, and file uploads are key functionality for a lot of web applications, so blocking them outright isn’t an option.
That takes out many of the traditional defenses. What remains is knowing where your critical web applications are. Identify them, starting with the admin interfaces on your firewalls and other network equipment, and restrict access to them. Palo Alto’s own guidance for PHP2Shell was to remove internet access to management interfaces and limit them to trusted internal IP addresses. If you rely on a web application firewall, configure it properly and understand how much protection it does and doesn’t provide.
Risk Was the Hardest Call
The clearest lesson from PHP2Shell is how difficult it can be to accurately determine the risk from a vulnerability like this. Prioritization is always a big issue for defenders, particularly with today’s flood of vulnerabilities, and this one was hard to read from every angle.
The scores told different stories. The PHP project rated both flaws 9.2 under CVSS 4.0, and Fortinet’s PSIRT listed 9.8 under CVSS v3. Palo Alto initially scored the issue 8.7 for its own products and listed it at moderate urgency, raised it to 9.2 Critical with the highest urgency on October 27, and confirmed exploitation a week after its advisory went out. Many public proofs of concept didn’t work, so a flood of attempts in your logs said little about whether a given server was at risk. And PHP core developer Gina Peter Banyard warned that one early workaround, turning off session upload progress, was not enough on its own.
We are adding these lessons to SEC522: how to assess risk, how protections like ASLR reduce it, and why appliances remain softer targets.
Defending What Runs on HTTP
I will be teaching SEC522: Application Security: Securing Web Applications, APIs, and Microservices at SANS Cyber Defense Initiative 2026, December 14-18 in Washington, D.C., and Live Online. The course starts with how HTTP works, including HTTP/2 and HTTP/3, and covers file upload handling, API security, and AI endpoints across 17 hands-on labs and a Defend the Flag capstone. It is built for the people who defend web applications, whether or not they wrote them.
Register for SEC522 at SANS Cyber Defense Initiative 2026.
Note: This article has been written and contributed by Dr. Johannes Ullrich, Dean of Research at SANS Technology Institute.





