Loading vulnerability…
CVE-2026-31908 is a header injection vulnerability in Apache APISIX affecting the forward-auth plugin. Under certain plugin configurations, an attacker can inject malicious HTTP headers. The issue affects Apache APISIX versions 2.12.0 through 3.15.0. Based on the available information, the flaw is tied to how the forward-auth plugin handles or forwards header data in specific configurations, enabling attacker-controlled header injection into downstream processing or requests.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
This repository is a small standalone Python proof-of-concept exploit for CVE-2026-31908, described as an Apache APISIX header injection issue affecting versions 2.12.0 through 3.15.0. The repo contains one executable script (CVE-2026-31908.py), a README, a LICENSE, and a minimal requirements.txt listing requests. The main script is structured into four functional stages: banner output, target fingerprinting, vulnerability testing, and exploitation. In check_vulnerability(), it sends a GET request to the supplied base URL, inspects the HTTP Server header for the string 'APISIX', and probes likely APISIX admin API paths (/apisix/admin/routes, /apisix/admin/services, /apisix/admin/upstreams) to infer APISIX deployment. In test_header_injection(), it sends several CRLF-based header injection payloads to a configurable protected path (default /protected), using manipulated Authorization, X-Forwarded-For, and Host headers, and treats unusual HTTP responses such as 400/500/502 as signs of possible vulnerability. In exploit_bypass_auth(), it attempts practical bypasses by injecting trust/admin/internal-use headers intended to trick upstream auth logic, such as localhost spoofing, auth-status success, admin role assignment, and internal service access markers. If a request returns HTTP 200, the script reports success and prints up to 500 bytes of the response body. This is not merely a detector: it actively sends exploit-style payloads intended to gain unauthorized access to protected resources. However, it is still best classified as an operational PoC rather than a highly weaponized tool, because payloads are hardcoded and the exploit logic is relatively simple. The attack vector is network/web-based, with the target being an HTTP service fronted by Apache APISIX. The most fingerprintable target-specific paths are the APISIX admin API endpoints and the default protected resource path /protected.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
3 sources tracked across advisories and community write-ups. News coverage will land here when it surfaces.
No news coverage yet. Advisories and community discussion only.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.