Executive Summary
CVE-2026-125283 lets a remote attacker trigger a segmentation fault in an NGINX worker process when the server is configured with the try_files and alias directives alongside a map variable, because specific combinations of regular expression captures in map and try_files fault the worker. The physical criticality is indirect but real: NGINX frequently sits in front of HMI web interfaces, historian dashboards, and remote access portals in modern plant and utility deployments, and repeated worker termination degrades or denies operator visibility into the process.
Technical Exposure Breakdown
This is a data plane memory safety defect, not a control plane exposure. According to the grounding data, there is no control plane impact, which means configuration management and the management API are not the attack surface. The fault occurs at request handling time inside a worker process. The condition set is narrow but not exotic. An operator must be running a configuration that combines try_files and alias with a map variable, and the attacker must send requests that produce the specific regex capture combinations that trip the segfault. The CVSS score is 5.3, consistent with an availability impact against a single component rather than full compromise.
The affected product spread is wide. Per the grounding data, NGINX Open Source is listed across ranges 0.9.6 through 1.30.7 and 1.31.0 through 1.31.9. NGINX Plus is listed across 37.0.0 through 37.0.8.1 and 37.1.0 through 37.3.0.1. Downstream products built on the same core are also in scope: NGINX Ingress Controller across several ranges including 5.0.0 through 5.8.0 and the 2026-lts-r1 through 2026-lts-r11 line, NGINX Gateway Fabric across 1.5.0 through 1.6.2 and 2.0.0 through 2.9.0, NGINX Instance Manager 2.19.0 through 2.25.0, and the F5 WAF and DoS components across their listed ranges. Patch status is unknown in the grounding data, so operators should not assume a fixed build exists for their specific deployment line until the vendor confirms one.
OT Impact and Compliance Risk
The physical failure mode is loss of view and loss of remote access, not loss of control. A worker crash does not corrupt PLC logic or send rogue setpoints. What it does is interrupt the web tier that operators and engineers rely on for monitoring and remote engineering sessions. In a control room that has consolidated multiple HMI front ends behind a single NGINX reverse proxy, a sustained crash loop can blind an operator at the worst possible moment.
For IEC 62443, this maps to availability requirements on the service and network layers, and it reinforces the zone and conduit argument that a web proxy facing anything beyond a tightly controlled zone is a conduit that must be hardened and monitored. For NERC CIP registered entities, a crashing proxy in an Electronic Security Perimeter affects CIP-007 system security management and CIP-005 remote access controls if the proxy fronts interactive remote access. For water and wastewater utilities under AWIA 2018, any component that mediates operator access to treatment control counts toward the risk and resilience assessment. For pipeline and rail operators under TSA SD-02C, availability of remote access and monitoring paths is a defined control objective.
Compensating Controls
Do not run an active scanner against production OT web tiers to find this. Active scanning of industrial components can brick devices and can itself trigger the exact fault class this CVE describes. Inventory passively from configuration files instead.
- Audit NGINX configurations for the specific pattern:
try_filesandaliasdirectives used together with amapvariable that performs regex capture. Where that pattern is not required, refactor it out. Removing the vulnerable configuration combination removes the attack surface regardless of version. - Place the proxy behind a filtering layer that normalizes request paths before they reach NGINX, stripping or rejecting the regex capture inputs that drive the fault.
- Deploy a Suricata rule concept at the conduit that alerts on anomalous URI patterns targeting the affected location blocks, keyed to the specific path prefixes handled by your
aliasandtry_filesdirectives. Treat this as detection first, then move to blocking once you confirm no false positives against legitimate operator traffic. - Configure worker auto restart and rate limiting so that a crash loop does not translate into sustained denial. This is mitigation of the symptom, not the defect.
- Restrict exposure of the proxy to known operator source ranges and a jump host, reducing the population that can reach the vulnerable code path.
BreachSpider tracks vulnerable NGINX and F5 component versions across OT deployments and flags configuration exposure as vendor patch status develops.
