Published on 31 July 2026
A single, unauthenticated HTTP request can be enough to crash an NGINX web server – or worse: to execute arbitrary malicious code on the server. That is exactly what CVE-2026-42533 makes possible, a critical vulnerability in the world's most widely used web server. Since 28 July 2026, a public, fully functional exploit has existed that bypasses the system's memory randomisation (ASLR) and reliably achieves remote code execution. Anyone running NGINX should act now.
On 15 July 2026, F5 – the vendor behind NGINX – published security advisory K000162097 and made patches available for a critical vulnerability. It carries the identifier CVE-2026-42533 and received a CVSS v4.0 score of 9.2 (Critical) in the official CVE database NVD – one of the highest possible values.
The core of the problem: a heap buffer overflow – an error in which more data is written into a reserved memory area than it can hold, overwriting adjacent memory. The vulnerability was discovered by security researcher Stan Shaw (known as "cyberstan"), who responsibly reported it to the F5 Security Incident Response Team on 17 May 2026 – including full proof that it can be leveraged for remote code execution.
Especially explosive: on 28 July 2026, just 13 days after the patch, the team "DepthFirstDisclosures" published a complete exploit-chain PoC (proof-of-concept, i.e. a working demonstration of feasibility) on GitHub. This puts the attack code in everyone's hands. As of our research on 31 July 2026, no active exploitation "in the wild" has been confirmed yet and the vulnerability is not listed in the CISA KEV catalogue – however, given the public PoC, the risk must be rated as high.
To understand why this vulnerability is so tricky, it helps to look at how NGINX works. The web server assembles composed strings – for example in directives such as proxy_set_header, add_header or return – using a two-pass procedure:
Both runs read from the same memory area, which contains the results of regex groups (the so-called captures, i.e. the matches of regular expressions such as $1, $2). The problem arises when a map directive with a regular expression overwrites these captures between the two runs: the LEN pass reserved the buffer for the original, smaller value – but the VALUE pass then writes the longer, attacker-controlled value into it. The result is an out-of-bounds write, i.e. a write access beyond the buffer boundary.
The flaw also works the other way round: if the overwriting value is smaller than the original, an oversized buffer with uninitialised memory content is left behind – and this is returned in the server's response. In this way, attackers gain access to internal memory addresses and can defeat the ASLR protection (Address Space Layout Randomization, the random arrangement of memory areas).
This is precisely what discoverer Stan Shaw emphatically points out:
"A reader of the F5 advisory could reasonably conclude that on standard systems this is only a DoS. It is not. The bug ships the bypass itself. […] On a standard Ubuntu 24.04 build, a single unauthenticated GET request recovers the addresses a payload needs. Tested at 10-out-of-10 reliability on Ubuntu 24.04 with glibc 2.39 and ASLR fully enabled." – Stan Shaw, The Hacker News
The PoC by DepthFirstDisclosures combines three steps: first, a memory leak is used to recover the bas