Plex2Shell: four Plex Media Server bugs chained to unauthenticated RCE
Four Plex Media Server flaws (CVE-2026-96652/96654/96655/96656, CISA KEV) chain from SSRF to a malicious codec library and are being used to drop miners and proxyware. Checks, IOCs and the fix.
Plex2Shell is a chain of four Plex Media Server (PMS) bugs. None of them looks dramatic on its own: the CNA scores range from 4.3 to 7.2. Chained together, they take an unauthenticated attacker on the internet to code execution as the Plex service account. Researcher zmain published the write-up on 16 October 2026. Exploitation began the next day, Plex now calls it “actively and widely exploited,” and CISA added all four CVEs to KEV on 20 October. This is the third of three KEV entries I’m covering this week, after F5 BIG-IP APM and ScreenConnect.
Scope and ethics. This post explains each link in the chain at the design level, plus exposure checks, detection and remediation for servers you own. It contains no request sequences or payloads. The IOCs come from published sources and are here for hunting.
At a glance
| CVEs | CVE-2026-96652, CVE-2026-96654, CVE-2026-96655, CVE-2026-96656 |
| Severity | Per-CVE (CNA, via NVD/CISA CSAF), CVSS 3.1: 4.3, 6.5, 4.3, 7.2 High. The researcher rates the full chain 9.8 |
| CWE | CWE-918 SSRF (96652, 96655), CWE-84 encoded URI schemes (96654), CWE-73 external control of file path (96656) |
| Affected | Plex Media Server before 1.43.3.10861 |
| Fixed in | 1.43.3.10861 and later |
| Exploited? | Yes. CISA KEV, added 20 Oct 2026. Plex: “actively and widely exploited.” Miners and proxyware observed |
| Auth / interaction | None for the chain (individual CVEs list low or high privileges) / none |
Description: how the chain works
Plex Media Server is a self-hosted media server. It runs a web API on TCP 32400, a plugin (“agent”) framework for metadata, and a transcoder that re-encodes video on the fly. Many people expose it to the internet so they can stream away from home, often by forwarding 32400 directly.
The chain works because several internal components trust requests that come from the server itself. Each link gives the attacker a bit more of that trust. Based on zmain’s write-up, Plex’s advisory and the CVE records:
- Unauthenticated SSRF (CVE-2026-96652). The player “companion” endpoints under
/player/timelinecheck only that anX-Plex-Tokenheader is present, not that it’s valid. A client-suppliedprotocolvalue is then used, without validation, to build a callback URL. Result: anyone can make the server send requests on their behalf, including to itself on loopback. - Loopback pivot into the agent API (CVE-2026-96654). The plugin-management API trusts loopback callers without a token. In the agent search function (
searchOne), a guard is broken because a decorator discards the check’s return value, and an identifier is inserted into an internal URL without encoding. Through the SSRF, an attacker can call other plugins’ functions with their own parameters. - Owner token disclosure (CVE-2026-96655). A bundled metadata plugin can fetch arbitrary URLs, and the plugin framework attaches the server owner’s admin token to any URL starting with a loopback-style prefix. The prefix match is unbounded, so a hostname that merely starts with that string but resolves to the attacker’s IP receives the token. (See the note below on how this CVE is described.)
- File write and code execution (CVE-2026-96656). With the owner’s token, the attacker is effectively the admin. The
TranscoderH264Optionspreference is “appended verbatim to x264’s option string on every transcode.” That lets an admin write attacker-controlled bytes to a chosen path in the Codecs directory. PMS then loads every.sothere “without signature, execute bit, or symbol checks.” A shared library with a constructor runs as soon as it’s loaded, as theplexuser.
A note on CVE-2026-96655. The NVD/CNA text describes it as an SSRF where “an authenticated user” can make the server request arbitrary addresses via /video/:/transcode/universal. Plex’s own advisory calls it “disclosure of the server owner’s admin token.” These are two views of the same outbound-fetch weakness, and they don’t change the remediation. Plex’s mapping is the one I’ve used above.
The same release also fixed CVE-2026-96651, an authenticated path traversal in /system/agents/media/get that can read the PlexOnlineToken (CISA CSAF). It isn’t in KEV and isn’t part of the chain.
Impact
- Code execution as the PMS service account, plus everything that account can reach: media shares, NAS volumes, Docker host mounts and stored credentials. Plex says to treat these as potentially exposed.
- Owner account takeover. The stolen admin token gives full control of the server and access to shared libraries.
- Monetization so far: cryptominers and residential-proxy software. Plex names
myawesomeproxy,mylovelyminerandx12345. BoredHackerBlog’s honeypot captured XMRig andmyawesomeproxythe day after disclosure.
Which deployments are exposed
| Deployment | Status |
|---|---|
| PMS before 1.43.3.10861, port 32400 reachable from the internet | Highest risk. Plex: treat as possibly compromised |
| PMS before 1.43.3.10861, LAN only | Exposed to anyone on the LAN or anything that can relay to it |
| NAS packages and Docker images | Same bug. Plex warns package managers may lag, so update manually |
| PMS 1.43.3.10861 and later | Not affected (Plex) |
Prerequisites: network reachability to PMS and a pre-fix build. No valid credentials and no user interaction for the full chain. The server also needs outbound internet access so the owner token can leak to the attacker, which most home and NAS deployments allow.
Identifying affected systems
Check the version in Plex Web > Settings > General, or from the host:
# 1) Version from the local API (the identity endpoint reports the server version)
curl -s http://127.0.0.1:32400/identity | grep -o 'version="[^"]*"'
# 2) Package version on Debian/Ubuntu or RHEL-family hosts
dpkg -s plexmediaserver 2>/dev/null | grep '^Version' || rpm -q plexmediaserver
# 3) Containers: which images are running Plex?
docker ps --format '{{.Names}} {{.Image}}' | grep -i plex
# 4) Is 32400 published beyond localhost/LAN? Check listeners and your router's port forwards
ss -ltnp | grep 32400
Anything below 1.43.3.10861 is affected. Also check Settings > Remote Access: if it’s enabled or 32400 is port-forwarded, assume the server was reachable from the internet.
Safe validation in an isolated lab
You can check your exposure and detection without running the chain:
- Stand up a pre-fix PMS container and a 1.43.3.10861+ container on an isolated network with no route to the internet or production.
- Confirm your version checks and scanner flag only the old one.
- Baseline the Codecs directory and
Preferences.xmlon the clean install, then confirm your file-integrity monitoring alerts when you drop a harmless test file into the Codecs folder or change a transcoder preference through the normal UI. - Confirm your EDR records the process tree for a normal transcode (
Plex Media ServertoPlex Transcoder), so you know what an unexpected child looks like.
Detection and threat hunting
Data paths below are for the Linux package (/var/lib/plexmediaserver/Library/Application Support/Plex Media Server/). Adjust for Docker, NAS or Windows installs.
- Codecs directory. Unfamiliar
.so(or.dll) files inCodecs/that aren’t part of the standard codec set. This is Plex’s top indicator.P="/var/lib/plexmediaserver/Library/Application Support/Plex Media Server" find "$P/Codecs" -type f \( -name '*.so' -o -name '*.dll' \) -printf '%TY-%Tm-%Td %TH:%TM %u %p\n' | sort - Preferences. Transcoder settings you didn’t configure in
Preferences.xml, especiallyTranscoderH264Options. The researcher notes attackers may change it and then change it back, so check backups and logs too.grep -o 'TranscoderH264Options="[^"]*"' "$P/Preferences.xml" - Server logs (
$P/Logs/Plex Media Server*.log): requests to/player/timelinewhoseprotocolparameter carries a full URL or a loopback address,/system/agents/activity that doesn’t match your own admin work, transcode requests whose source is an external URL instead of a library item, and thePython-urllib/3.10user agent seen in honeypot exploitation. - Outbound connections from the Plex host to hostnames that start with a loopback-looking string (for example under
sslip.ioor similar wildcard-DNS services), toproxystream.st, or to mining pools. - Processes.
myawesomeproxy,mylovelyminer,x12345,xmrig, or any process running asplexthat isn’tPlex Media Server,Plex Transcoderor another Plex component. High CPU while idle is the cheap tell.# review anything here that isn't a Plex component; sort by CPU ps -u plex -o pid,etime,pcpu,cmd --sort=-pcpu
Published IOCs (BoredHackerBlog honeypot, 17 Oct 2026)
- Exploitation sources:
91.92.40.202,176.65.139.228,45.194.92.222 - Payload host:
217.60.194.116(serveddockerandx) - Domain:
proxystream.st - SHA256:
5712ef510481f15fe8b18fdd10b73026d57cb3d9d151878a692358b99a464441(docker)68938fb428f2bcfef4ede62f7b29cd0f15f37ed59e94462c099f2ca64e36baa7(x, UPX-packed)b20f39fc00d242e706b6c30367ad811c676e0575050a4ec2f30104b696944b49(x, unpacked)e546b10aee5edc754da4f9b0baace00392257121911f630145078eafad1e6bd9(myawesomeproxy)2d3c618b21a7e2f8bcd735da188e44a7609ff3f204832dd0781cd09e4388c9db(xmrig)
The researcher’s repository has a detection folder, and BoredHackerBlog published PCAPs for building network signatures (see references).
Remediation
- Update to Plex Media Server 1.43.3.10861 or later. NAS and Docker users may need to update manually.
- If the server was internet-reachable on an older build, assume compromise until you’ve checked. Plex recommends a rebuild from a clean install rather than reusing the old data directory. At minimum, clear the Codecs folder and review
Preferences.xmlfor transcoder settings you didn’t set. - Sign out of all devices in your Plex account and change your Plex password. This invalidates the stolen owner token.
- Rotate anything the Plex account could reach: NAS or SMB credentials, cloud storage keys, and any secrets on mounted volumes.
Interim mitigation (if you can’t patch today)
- Turn off Remote Access if you don’t need it, and stop forwarding 32400 directly. Use a VPN for remote viewing.
- Block outbound traffic from the Plex host to anything it doesn’t need. This breaks the token-exfiltration step, but it isn’t a fix.
- Monitor the Codecs directory and
Preferences.xmlwith file-integrity monitoring.
References
- Plex security update for PMS 1.43.2 and earlier
- zmain: Plex2Shell research write-up · code-zm/plex2shell detection folder
- BoredHackerBlog: Plex vulnerabilities exploited in the wild · honeypot PCAPs
- NVD: CVE-2026-96652 · CVE-2026-96654 · CVE-2026-96655 · CVE-2026-96656
- CISA CSAF VA-26-266-01 · CISA KEV catalog
Verify every version against the vendor advisory before acting. Published 06 Nov 2026.