F5 patches five critical NGINX flaws dubbed NGINX2Root; public exploit exists for one
The bugs can allow remote code execution in the web server, and researchers chained them with an unpatched Linux kernel flaw to reach root.
By Mr Elite Published 7 min read
- Severity
- High
- Status
- Reported by outlets
- Sources
- 1
- Last checked
The short version
- F5 patched five critical NGINX vulnerabilities, named NGINX2Root by Xint researchers, Help Net Security reported on 19 November 2026.
- No exploitation in the wild is known, and none were in CISA's KEV catalog as of 19 November.
- A public proof-of-concept exploit exists for CVE-2026-123007; Rapid7 describes it as constrained.
- Fixed releases are NGINX Open Source 1.30.7 and 1.31.9, and NGINX Plus 37.0.8.1 and 37.3.0.1.
- The Linux kernel flaw used to reach root had no CVE and no patch as of 18 November.
What happened
F5 has patched five critical vulnerabilities in NGINX, the widely used open-source web server, Help Net Security reported on 19 November 2026. Under certain conditions the flaws can allow remote code execution, meaning an attacker runs their own commands on the server.
Xint researchers named the set NGINX2Root. Two of the flaws, CVE-2026-123633 and CVE-2026-123007, were also found and reported independently by researchers at Rapid7, Wiz, Ikotas Labs, Palo Alto Networks and Metnir, according to the outlet.
Help Net Security said there is no evidence that any of the five has been exploited in the wild. None was listed in CISA's Known Exploited Vulnerabilities (KEV) catalog, a US government list of flaws attackers are known to be using, as of 19 November.
A public proof-of-concept exploit, a working demonstration of the attack, is already available for CVE-2026-123007. Rapid7's Stephen Fewer published a technical analysis and the exploit. Xint and Wiz have built exploits for all five flaws but have not released them.
F5 fixed five further NGINX flaws in the same releases. NGINX also sits inside other F5 products, including NGINX Instance Manager, NGINX Ingress Controller and NGINX Gateway Fabric. F5 has updated some of these and is still working on the rest, the outlet reported.
Why it happened
Four of the five flaws are use-after-free bugs, where a program keeps using a piece of memory after it has been released. CVE-2026-123633 is reachable when HTTP/2 is used and requests are forwarded to an upstream server. CVE-2026-123634 sits in NGINX's DNS resolver and needs the resolver, request-derived proxy_pass and mirror directives (with mirror_request_body off) configured together.
CVE-2026-121810 comes from closing a QUIC connection while an HTTP/3 request is running. CVE-2026-113425 is in the mail proxy's auth_http path. The fifth, CVE-2026-123007, also called NGINX2Shell, is an integer underflow in HTTP/2 flow control that leads to out-of-bounds reads and writes on the heap. It can crash an NGINX worker process and potentially allow code execution.
If address space layout randomisation (ASLR) is disabled or bypassed, any of the five can give code execution with the privileges of the NGINX worker process. ASLR shuffles where code and data sit in memory so attackers cannot predict addresses. Getting from there to root needs a sixth flaw, which Xint also found: a use-after-free in the Linux kernel's SCTP code that can become a double free. It had no CVE as of 18 November and was unpatched.
How this kind of failure works
Memory-safety bugs like use-after-free arise in languages such as C, where the programmer manages memory by hand. Picture a hotel that gives you room 214, then hands the same key to a new guest after you check out. If you still hold your old key, you can walk into a room that now belongs to someone else, and read or change their things.
In software, the 'new guest' can be data an attacker controls. If a program later uses the stale pointer, it may follow attacker-chosen instructions. Protections such as ASLR make this harder because the attacker must first learn where things are in memory. That is why the reporting stresses ASLR as a condition.
Exploit chains matter because a bug in a low-privilege process is a smaller problem than a bug that reaches the whole machine. A web server worker usually runs with limited rights. A second flaw, here in the kernel, can lift an attacker from that limited position to full control, and from a container onto its host. Each link looks less alarming alone than the chain does together.
Who is affected
Anyone running NGINX Open Source or NGINX Plus before the fixed releases is in scope, along with F5 products that embed NGINX. Exposure depends on configuration: for example, CVE-2026-123633 needs HTTP/2 with an upstream server, and CVE-2026-113425 concerns the mail proxy. Check your running version and which features you have enabled.
Rapid7's exploit for CVE-2026-123007 is constrained, according to Tenable researchers quoted by Help Net Security. It relies on a glibc memory-copy routine specific to x86-64 builds, needs roughly 128 gigabytes of HTTP/2 traffic, and must recover heap addresses from server responses to beat ASLR. Code runs in the worker process.
Tenable also said it is unlikely to work against servers behind HTTP/2-terminating proxies, or on connections too slow to deliver the data before NGINX's one-hour connection limit. These are limits on one public exploit, not a guarantee for the other four flaws. Xint has published instructions for investigating possible attacks via CVE-2026-123007.
How to defend against it
Tick each step as you check it. Your progress is saved on this device.
What to learn from this
The concept: Exploit chaining and defence in depth
A single vulnerability rarely gives an attacker everything. In NGINX2Root, the web server flaws give code execution as a limited worker, and a separate kernel flaw is needed for root. Defenders who rank bugs one at a time can miss how they combine.
Defence in depth means placing several independent barriers so that one failure does not hand over the whole system. ASLR, low-privilege workers, container isolation and fewer exposed services each raise the cost of a chain. Remove any one and the others still matter.
The same pattern appears in phishing followed by privilege escalation, and in a cloud misconfiguration paired with a stolen token. Ask what an attacker could do next from each foothold, not only how bad each bug is alone.
Assess vulnerabilities as links in a chain, and keep several independent barriers so one flaw does not give away the host.
Quick check
Sign in to earn 10 XP for a right answer
Why did researchers need a Linux kernel flaw in addition to the NGINX bugs to get root?
Per the reporting, the NGINX flaws yield the worker's privileges. Reaching root required a separate flaw in the kernel's SCTP code.
Timeline
- Before 19 November 2026 (exact dates not given) Xint researchers, and for two flaws Rapid7, Wiz, Ikotas Labs, Palo Alto Networks and Metnir, report the bugs.
- 18 November 2026 The Linux kernel SCTP flaw used for the root step has no CVE and is unpatched.
- 19 November 2026 F5 patches five critical NGINX flaws; Help Net Security reports no known exploitation and no KEV listing.
- 19 November 2026 A public proof-of-concept for CVE-2026-123007 is available, published by Rapid7's Stephen Fewer.
What is not known yet
- Whether and when the unpatched Linux kernel SCTP flaw will be fixed and given a CVE.
- When Xint and Wiz will release their exploits for the five flaws, if at all.
- When F5 will finish updating the remaining products that embed NGINX.
- Whether any of the flaws is being exploited in the wild; none has been reported.
Questions people ask
What is NGINX2Root?
It is the name Xint researchers gave to five critical NGINX vulnerabilities that F5 has patched. They can allow remote code execution under certain conditions. The name refers to the researchers' chain that reached root using a separate Linux kernel flaw.
Are the NGINX flaws being exploited?
Help Net Security reported no evidence of exploitation in the wild, and none was in CISA's KEV catalog as of 19 November. A public proof-of-concept exists for CVE-2026-123007. That means the picture could change quickly.
Which NGINX versions fix the flaws?
NGINX Open Source 1.30.7 and 1.31.9, and NGINX Plus 37.0.8.1 and 37.3.0.1, contain the fixes. After upgrading, restart NGINX, because a configuration reload does not replace the running executable. Distro package users should check for backported fixes.
Is the public exploit easy to use?
Rapid7 describes it as constrained, according to Tenable. It needs an x86-64 build with a specific glibc routine, roughly 128 gigabytes of HTTP/2 traffic, and a way to defeat ASLR. It is unlikely to work behind HTTP/2-terminating proxies.
Sources
- Help Net Security: NGINX2Root: F5 patches five critical NGINX vulnerabilities