QNAP has warned that attackers are breaking into network-attached storage (NAS) devices through a chain of four Plex Media Server flaws known as Plex2Shell. The company says it has received reports of suspected compromises and that its support team is helping affected customers. The flaws sit in the Plex application, not in QNAP’s operating systems, but a vulnerable Plex server exposed to the internet can hand an attacker code execution on a device that often holds backups, project files and shared credentials.
This matters beyond home media libraries. NAS boxes are common in engineering offices, plants and small industrial sites, where they hold PLC project files, HMI backups, vendor documentation and system images. This article separates what is confirmed from what is not, and ends with a practical response checklist.
Key facts at a glance
- What QNAP confirmed: it received reports that Plex Media Server on customer NAS devices appears to have been compromised via Plex2Shell. The wording is “suspected” compromise; QNAP has not published a count of affected devices (BleepingComputer).
- Vulnerabilities: CVE-2026-96652, CVE-2026-96654, CVE-2026-96655 and CVE-2026-96656, all in Plex Media Server.
- Fixed version: Plex Media Server 1.43.3.10861 or later.
- Not affected: QTS, QuTS hero and other QNAP operating systems, per QNAP PSIRT.
- Government action: CISA added all four CVEs to its Known Exploited Vulnerabilities (KEV) catalog on October 20, 2026.
- Attribution: none. No threat actor has been publicly identified.
What happened
On November 6, 2026, QNAP published a security notice saying some users reported that the Plex Media Server on their NAS appeared compromised through the Plex2Shell vulnerabilities, and that QNAP Technical Support was assisting them. QNAP’s PSIRT stated that the bugs exist only in Plex Media Server.
QNAP’s guidance is blunt. If a vulnerable Plex server was reachable from the public internet, treat the NAS and its content as compromised. Updating Plex closes the door but does not remove anything an attacker already planted.
The QNAP notice is the latest step in a campaign that began weeks earlier, when a public exploit appeared (see the timeline below).
Timeline of the Plex2Shell campaign
| Date (2026) | Event |
| July 27 | Researcher Zach Main reports the flaws to Plex |
| Early August | Plex confirms the issues and ships 1.43.3.10861 (sources differ on whether the release date is August 3 or 4) |
| September 1 | Plex urges server owners to update, but the advisory carries no CVE IDs |
| September 23 | CVE records are published by CISA acting as CNA |
| October 16 | Main publishes a technical write-up and proof-of-concept code |
| October 17 | First in-the-wild exploitation is observed on a researcher’s honeypot, about a day after the PoC release |
| October 19 | Plex updates its advisory with the CVE list and indicators of compromise, saying the flaws are being actively and widely exploited |
| October 20 | CISA adds all four CVEs to the KEV catalog |
| November 6 | QNAP reports suspected NAS compromises and issues remediation guidance |
Sources: SecurityWeek, BleepingComputer on CISA’s listing, BetaNews on Plex’s update, CVE record for CVE-2026-96652.
The patch existed for roughly three months before mass exploitation began. In September, the nonprofit Shadowserver reported more than 36,000 internet-exposed Plex servers still unpatched (BleepingComputer). That number is from September; current exposure is not confirmed here.
Root cause: four flaws, one unauthenticated path to code execution
Plex2Shell chains four weaknesses so that an attacker with network access to the server, and no credentials, can run code as the Plex service user. According to the researcher’s public repository and CVE descriptions, the primitives are:
| CVE | Weakness (per researcher’s description) |
| CVE-2026-96652 | Unauthenticated server-side request forgery (SSRF) in the player timeline function |
| CVE-2026-96654 | Pivot through requests the server sends to itself, abusing a non-functioning authentication check and an injection in an agent search function |
| CVE-2026-96655 | SSRF in the transcoder, reachable with owner rights, plus an arbitrary file write |
| CVE-2026-96656 | Code execution because the server loads unsigned codec libraries from its Codecs directory |
Source: researcher’s plex2shell repository. CVE-2026-96656 is the most severe individual flaw; one report lists it at CVSS 7.2 (v3.1) and 8.6 (v4.0) (WWWhatsNew). The researcher scores the full chain at CVSS 3.1 9.8, which is a self-assessment and not an official rating. Individual CVEs in the chain score lower on their own, which is a useful reminder that chains can be far worse than their parts.
In plain terms: the first flaws let an outsider make the server talk to itself and obtain owner-level access, a transcoder option is then used to write a file into the Codecs folder, and the server loads that file as a library. The affected versions are 1.43.2 and earlier per Plex’s advisory; the researcher’s repository lists versions up to 1.43.3.10828 as vulnerable, with the fix in 1.43.3.10861.
Attack methodology
Initial access. Internet-exposed Plex Media Server, typically via forwarded port 32400. No user interaction or credentials are needed (Field Effect).
Execution and persistence. Per reporting on the chain, the malicious library sits in the Codecs folder and in a stored preference, so it can run again whenever transcoding occurs. A service restart is therefore not a cleanup.
Payloads observed (honeypot data). A researcher running a vulnerable honeypot saw an XMRig-based cryptocurrency miner, then a program called myawesomeproxy, which behaves as proxy software and displays a notice about use of the host’s internet traffic (SecurityWeek). The researcher noted that residential-IP proxies are valuable to criminals for reaching corporate VPNs and email. That is the researcher’s reasoning, not confirmed attacker intent. The researcher also cautioned that published indicators reflect only what the honeypot saw.
What is not documented. Public reporting does not describe lateral movement, privilege escalation or data theft on real QNAP devices. The honeypot was the only documented compromise in early reporting, and QNAP has not shared technical details of the customer cases.
Threat actors and attribution
There is no confirmed attribution. Security vendors reviewing the campaign say it has no definitive link to a known APT or botnet (Rescana, SecNews). CISA lists ransomware use of these CVEs as “unknown.” The observed payloads, mining and proxying, point to financially motivated opportunists, but that is an inference, and a public PoC means many unrelated actors can run the same exploit.
Impact assessment
Confirmed: compromised Plex servers have been used for cryptomining and proxying; QNAP has received multiple reports of suspected NAS compromise; CISA treats the flaws as exploited in the wild.
Likely, based on how the flaw works (analysis, not confirmed): code runs as the Plex service account, so anything that account can reach is in scope. Plex’s own guidance says to treat mounted shares and locally stored credentials as potentially exposed after a confirmed compromise (BetaNews). Direct effects such as high CPU load degrade the NAS’s performance for its real job.
Unknown: number of compromised QNAP devices, whether any data was stolen or encrypted, any industrial or enterprise victims, and regulatory consequences. QNAP has not reported financial figures, and none should be assumed.
IT, OT and IoT implications
Plex2Shell is not an ICS vulnerability, and no industrial victims have been reported. The relevance to OT security is about where such devices live:
- NAS as shadow infrastructure. Engineering workstations, historians and maintenance teams often rely on NAS for project backups, firmware images and documentation. A media app installed on that same box, perhaps by one employee for convenience, widens the attack surface of a system that holds crown-jewel files.
- Service-account reach. An attacker who inherits the Plex service account inherits whatever shares it was given. Over-broad permissions turn a media-server bug into a data-store compromise.
- Residential and corporate proxying. A compromised device can relay attacker traffic through your network address, which can bypass IP-based trust and implicate your organization in malicious activity.
- Compromise persists after patching. Teams that “patched and moved on” may still be hosting an implant.
- Same-week OT-adjacent signal. CISA’s October 20 KEV update also listed CVE-2026-110936, an uncontrolled search path flaw in Siemens Software Center, for which Siemens published indicators of compromise and urged an update to 4.0.9.8 (BleepingComputer). The two are unrelated, but they show attackers working both consumer-grade and industrial software at once.
Response and recovery
QNAP has issued a security notice and a step-by-step remediation FAQ, says the fixed version is available in the App Center, and offers to analyze logs from suspected compromises submitted through its technical support portal. Its steps for exposed, vulnerable servers: check for signs of compromise using the bulletin, perform a clean reinstall of QTS or QuTS hero, and reset Plex and NAS credentials. Users who do not need remote access should disable it and stop forwarding port 32400; those who do should use a VPN.
Plex shipped the fix in early August, published its advisory on September 1, and on October 19 added the CVE list and indicators of compromise.
CISA set a three-day remediation window for federal agencies (October 23 for some of the CVEs, with later dates reported for others).
Recovery status: ongoing. QNAP has not announced a final count or declared the incident closed.
Indicators of compromise to check
Reported indicators (from Plex’s updated advisory and researcher reporting; not exhaustive):
- Child processes launched by the Plex Media Server process
- Processes named myawesomeproxy, mylovelyminer or x12345
- Command-line arguments referencing a rig ID or donate level (typical of miners)
- Outbound connections to the domain proxystream.st
- Unfamiliar .so (Linux) or .dll (Windows) files in the Plex Codecs folder
- Sustained high CPU while nothing is streaming
Absence of these indicators does not prove a clean system, since attackers can rename payloads. Detection rules (YARA) for the dropped files are published in the researcher’s repository.
Security recommendations
- Find every Plex instance. Inventory NAS apps, Docker containers and workstations. Anything on engineering or backup storage should be questioned.
- Update to 1.43.3.10861 or later, but check for compromise first if the server was ever internet-exposed. Preserve logs.
- Assume compromise if it was exposed and old. Follow QNAP’s clean reinstall, then rotate Plex, NAS admin and any credentials stored on or reachable from the device. Plex also advises signing out of all devices.
- Verify backups. Restore only from copies you can show predate the exposure.
- Close port 32400 to the internet. Use a VPN or zero-trust access for remote use.
- Apply least privilege. Give service accounts only the shares they need; keep media apps off systems that hold engineering or operational data.
- Segment. Keep NAS devices that store OT artifacts in a controlled zone, not directly reachable from the internet, and restrict what they can initiate outbound.
- Monitor egress. Alert on new outbound destinations from storage devices and on sustained CPU anomalies.
- Treat CISA’s KEV catalog as a patch-priority feed for all internet-facing software, including “non-business” apps.
- Practice the playbook. Run a tabletop exercise on a “patched but possibly already breached” device.