Summary
Koel: Incomplete fix for CVE-2026-47260, systemic SSRF in podcast & radio fetch paths
The fix for CVE-2026-47260 (v9.3.5) added an initial isSafeUrl() check to several fetchers (synchronizeEpisodes, getStreamableUrl, AddRadioStation, EpisodePlayable), but the redirect-target validation, the per-hop Guzzle on_redirect callback added in follow-up commit be1e867, was applied to only one path, EpisodePlayable. Every other server-side fetcher therefore has only the initial check, which an HTTP 302 redirect to an internal address bypasses, or no check at all. DNS rebinding (validation and connection resolve DNS separately, with no IP pinning) bypasses the initial check on every path.
An authenticated, non-admin user can thus cause the Koel server to issue requests to arbitrary internal / cloud-metadata endpoints (SSRF) by supplying a URL on an attacker-controlled host that 302-redirects to an internal address.
Note: commit be1e867 shows the redirect-based SSRF vector was recognised, but the redirect defense was applied to a single call site rather than generalised, so the class survives in the sibling paths below.
Details, Root cause
App\Helpers\Network::isPublicHost() / isSafeUrl() perform a point-in-time host check with no pinning of the resolved IP, and per-redirect-hop re-validation exists only in App\Values\Podcast\EpisodePlayable (the on_redirect callback from commit be1e867). Consequently every other fetcher is exposed to (1) redirect SSRF, initial URL passes isSafeUrl, then the HTTP client follows a cross-host 302 to an internal target without re-validating the hop; and (2) DNS rebinding (TOCTOU), isPublicHost resolves DNS at validation, the HTTP client resolves again at connect time.
Affected paths (all reachable by any authenticated user)
| # | Location | Issue |
|---|---|---|
| 1 | PhanAn\Poddle\Poddle::fromUrl() → Http::timeout()->get($url) (used by PodcastService::addPodcast/refreshPodcast) |
Plain Http::get, follows redirects, no per-hop validation; refreshPodcast does not re-run isSafeUrl at all |
| 2 | PodcastService::getStreamableUrl() (PodcastService.php:244/251) |
Has the initial isSafeUrl() (line 244) but the request uses ALLOW_REDIRECTS => ['track_redirects' => true] with no on_redirect → 302 to internal is followed. Called at episode stream time via PodcastStreamerAdapter. Also DNS-rebinding-exposed |
| 3 | PodcastService::isPodcastObsolete() (:221) Http::head($podcast->url) |
No isSafeUrl, no redirect validation |
| 4 | App\Rules\HasAudioContentType (:45/:54) Http::head/Http::get |
Self-documented "use after SafeUrl"; ordering-dependent, no own validation, no per-hop check. Extends the surface to the internet-radio feature (RadioStationStore/UpdateRequest) |
| 5 | App\Rules\SafeUrl validator (:52/:56) |
Follows redirects, validates only the final effective host, intermediate-hop requests still fire |
Reachable via the native API (apiResource podcasts, radio/stations; PodcastController::store has no authorization check, only #[DisabledInDemo]) and the Subsonic API (createPodcastChannel, createInternetRadioStation, refreshPodcasts).
PoC
A mechanism PoC that runs the exact Guzzle/Laravel-Http call shapes Koel uses (attacker-redirect server + internal-target listener on loopback), verified on PHP 8.2 + Guzzle 7:
isPublicHost('127.0.0.1') = false # a per-hop check WOULD block this
Case1 Poddle::fromUrl -> [VULNERABLE] leaked INTERNAL-SECRET-TOKEN
Case2 getStreamableUrl -> [VULNERABLE] leaked INTERNAL-SECRET-TOKEN
Case3 EpisodePlayable -> [BLOCKED] UnsafeUrl on redirect
internal_hits.log: 2 hits # internal service actually reached by Case1 + Case2
Case1/Case2 reaching the internal target while Case3 (the fixed path) blocks under identical conditions demonstrates the incomplete remediation. The full PoC kit (poc.php, attacker_router.php, internal_router.php) is available on request.
End-to-end on a real instance: an authenticated user POST /api/podcasts (or Subsonic createPodcastChannel) with a feed URL on an attacker host that returns 302 Location: http://169.254.169.254/latest/meta-data/... (or http://127.0.0.1:<port>/); the server follows it. The response is reflected back via parsed podcast fields / getStreamableUrl when the internal endpoint returns Access-Control-Allow-Origin: *; otherwise blind SSRF via status/timing.
Impact
Authenticated (any user) SSRF: access to cloud instance metadata (IAM credentials on IMDSv1), internal-only admin panels, and internal network service probing, from the Koel server's network position. Same threat model as CVE-2026-47260.
Attack scenario (fully remote, no user interaction): the only precondition is a single low-privilege account. On AWS/GCP/Azure-hosted instances, redirecting to the metadata IP and reflecting the body discloses temporary IAM credentials → cloud-account pivot. (AWS IMDSv2's token-via-PUT is not reachable through a simple GET-redirect SSRF; IMDSv1 instances are fully exposed.) "Koel only runs on an internal/trusted network" does not reduce the risk, the bug makes the Koel server itself the attacker's pivot into that trusted network and cloud control plane.
Suggested severity: CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N (7.1); lower where exploitation is blind.
Untrusted input controls the target URL of a server-initiated request, which may reach internal services not otherwise accessible from outside. Typical impact: access to internal metadata services, internal APIs, or cloud credentials.
CVE-2026-54491 has a CVSS score of 7.1 (High). The vector is network-reachable, low privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (9.7.1); upgrading removes the vulnerable code path.
Affected versions
Security releases
Kodem intelligence
Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.
Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.
Already deployed Kodem?
See it in your environmentNew to Kodem? Get a demo →Remediation advice
Do not fix per-call-site. Centralize: route all outbound HTTP through a shared Guzzle handler/middleware that, on every connection and every redirect hop, resolves the target and rejects private/reserved IPs, and pins the validated IP for the actual connection (defeats DNS rebinding). Apply to EpisodePlayable, getStreamableUrl, Poddle::fromUrl usage, isPodcastObsolete, HasAudioContentType, and the SafeUrl rule.
Frequently Asked Questions
- What is CVE-2026-54491? CVE-2026-54491 is a high-severity server-side request forgery (SSRF) vulnerability in phanan/koel (composer), affecting versions <= 9.7.0. It is fixed in 9.7.1. Untrusted input controls the target URL of a server-initiated request, which may reach internal services not otherwise accessible from outside.
- How severe is CVE-2026-54491? CVE-2026-54491 has a CVSS score of 7.1 (High). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
- Which versions of phanan/koel are affected by CVE-2026-54491? phanan/koel (composer) versions <= 9.7.0 is affected.
- Is there a fix for CVE-2026-54491? Yes. CVE-2026-54491 is fixed in 9.7.1. Upgrade to this version or later.
- Is CVE-2026-54491 exploitable, and should I be worried? Whether CVE-2026-54491 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
- What actually determines whether CVE-2026-54491 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
- How do I fix CVE-2026-54491? Upgrade
phanan/koelto 9.7.1 or later.