guzzlehttp/psr7 EUVD-2026-36239

Exploitation requires all of the following conditions to be true simultaneously: (1) the application calls one of the affected inbound-parsing functions - GuzzleHttp\Psr7\Message::parseRequest(), the legacy GuzzleHttp\Psr7\parse_request(), GuzzleHttp\Psr7\ServerRequest::fromGlobals(), or GuzzleHttp\Psr7\ServerRequest::getUriFromGlobals() - on attacker-supplied HTTP data or server variables; AND (2) the application subsequently uses the resulting PSR-7 URI getHost() value for routing, allow-list enforcement, credential selection, or request forwarding decisions. Applications using guzzlehttp/psr7 only through the standard Guzzle outbound HTTP client API are explicitly confirmed not vulnerable by the vendor advisory. No attacker authentication is required, but exploitation is structurally limited to forwarding proxies, API gateways, or server-side request routing middleware. Standard PHP applications that receive requests via a web server and use psr7 solely for outbound calls are not in scope.

The NVD-assigned CVSS 3.1 score is 5.3 Medium (AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N). The network attack vector and absence of required privileges are consistent with the description - the attack is delivered via a standard HTTP request with no authentication. However, the practical exploitation space is narrower than the raw vector implies: only applications in specific inbound-parsing-plus-forwarding configurations are vulnerable; applications using guzzlehttp/psr7 exclusively through the Guzzle outbound client are explicitly confirmed not affected by the vendor advisory. No EPSS score was included in the source data, no CISA KEV listing exists, and no proof-of-concept exploit has been identified, all of which reduce urgency relative to the CVSS score alone. The assigned C:L/I:N metrics arguably understate the forwarding scenario - credential disclosure to an attacker-controlled host and request misdirection both carry integrity implications - but the scope is ungranged (S:U). Real-world priority is elevated specifically for API gateways, reverse proxies, or multi-tenant request routing middleware built on psr7's server request functions; for applications using only the outbound Guzzle client the effective risk is negligible.

An attacker targets a PHP application built on guzzlehttp/psr7 that acts as an API gateway or forwarding proxy. The attacker sends an HTTP request with Host: internal-api.corp@attacker.example; the psr7 library constructs a URI with getHost() returning attacker.example, which passes an allow-list check that only examines the parsed URI host. The gateway then forwards the request - including Authorization headers or session tokens - to the attacker-controlled host instead of the intended internal service.