Your server’s security software is supposed to be the last thing you worry about. It runs as root, it watches for intruders, and it updates itself quietly in the background. So when the update channel itself gets poisoned, the usual rules flip upside down: the thing protecting your server becomes the thing attacking it. That is exactly what happened in the cPGuard breach on 2 November 2026, when a fake cPGuard update ran as root on about 2,950 Linux hosting servers and copied their most sensitive secrets back to an attacker.
If you run cPanel, DirectAdmin or Plesk on a VPS, there is a real chance cPGuard is installed on your box — it is a popular, cheap malware scanner that a lot of small hosts and server admins install alongside their control panel. And if your server checked for updates between 02:08 and 03:20 UTC on 2 November, you should assume the attacker has your root password hashes, your root SSH keys and your MySQL root password. Updating cPGuard removes the backdoor, but it does not un-steal those credentials. This guide walks through what happened, how to check whether your VPS was hit, and the exact steps to lock it down again — most of it takes under an hour per server.
- What Happened in the cPGuard Breach
- What the Fake Update Stole From Your Server
- Was Your VPS Affected? How to Check in 5 Minutes
- Step 1: Update cPGuard and Remove the Backdoor
- Step 2: Rotate Root Passwords and SSH Keys
- Step 3: Change the MySQL Root Password
- Step 4: Rotate Secrets From Shell History and Cloud Keys
- Step 5: Check Every Server That Trusted Your Keys
- On a Managed VPS? Ask Your Host These 5 Questions
- When to Rebuild the Server From Scratch
- Why Supply-Chain Attacks Are the New Normal for VPS Owners
- Frequently Asked Questions
- The Bottom Line
What Happened in the cPGuard Breach
cPGuard is a malware scanner for Linux hosting servers, made by a small company called OPSSHIELD. It runs with root privileges, which is the whole point — it needs deep access to scan files, quarantine malware and block attackers. On 1–2 November 2026, an attacker turned that trust against it.
According to OPSSHIELD’s own disclosure, published on 3 November, the attacker started by registering as a customer. They signed up for a free cPGuard trial under a false identity, installed it on their own server, and studied how the agent talked to the licensing system. cPGuard actually flagged and banned them twice that evening — and the attacker got around it anyway. Late on 1 November they sent crafted requests to the billing portal’s callback pages and gained the ability to run commands on OPSSHIELD’s billing and licensing server.
From there it moved fast. The attacker stole the access key OPSSHIELD uses to publish updates, uploaded a fake cPGuard update to the company’s update storage, and for about 70 minutes on 2 November, customer servers that checked for updates installed it. Because cPGuard runs as root, the fake update ran as root too. Here is the timeline, taken from the vendor’s disclosure:
| When (UTC) | What happened |
|---|---|
| 1 Nov, 15:38 | Attacker begins probing the OPSSHIELD billing portal |
| 1 Nov, 16:34 | Attacker registers a customer account under a false identity and orders a free cPGuard trial |
| 1 Nov, 21:17 | Attacker installs the trial on their own server to study how the agent talks to the licensing system |
| 1 Nov, 21:40–21:51 | cPGuard flags and bans the attacker twice |
| 1 Nov, 22:50 | Attacker sends crafted requests to the billing portal’s callback pages |
| 1 Nov, 22:58 | Attacker gains the ability to run commands on the billing and licensing server |
| 2 Nov, 02:04–02:09 | Attacker uploads a fake cPGuard update using a stolen access key |
| 2 Nov, 02:08–03:20 | cPGuard agents install the fake update; it runs as root and copies sensitive files back to the attacker |
| 2 Nov, 02:13 | Attacker makes copies of three of OPSSHIELD’s cloud servers |
| 2 Nov | Stolen update key revoked; clean update v5.88.26 published to remove the attacker’s tools |
| 3 Nov | Entry point traced; disclosure published |
The speed is the scary part. From first probe to fake update took about ten hours. OPSSHIELD itself says attackers now use AI tools to scan for weak points, read unfamiliar code and write custom tools in hours instead of weeks — and this attack is a textbook example. A small vendor, a billing portal nobody was watching closely, and a trusted update channel did the rest.
This is a supply-chain attack, not a vulnerability in your server. Your firewall did not fail, your passwords were not guessed, and no plugin on your site was exploited. The malicious code arrived through the same mechanism you normally trust for security updates. That is why the response is different from a normal patch: updating fixes the software, but everything the fake update copied is still in the attacker’s hands until you rotate it.
What the Fake Update Stole From Your Server
The fake update was a short script, not a long-term implant. It ran once, collected sensitive files, and sent them to a server the attacker controlled. On the servers that ran it, it read:
| What was copied | Why it matters |
|---|---|
/etc/shadow and /etc/passwd |
Local password hashes for every user on the server — crackable offline at the attacker’s leisure |
/root/.ssh |
Root SSH private keys, authorized_keys and known_hosts — enough to log in as root wherever those keys are trusted |
/root/.my.cnf |
MySQL root password, where present — full access to every database on the box |
| Root shell history | Often contains typed passwords, tokens and API keys from past admin sessions |
/root/.s3cfg |
Cloud storage keys, on a small number of servers — access to backups and object storage |
It also left two things behind: a hidden root shell at /var/tmp/.cpg91511_sh and files in /var/tmp/.i93677/, plus a scheduled task at /etc/cron.d/i93677 to keep the backdoor alive. The clean update v5.88.26 removes the hidden shell, and a follow-up update removes the scheduled task — but you should verify both are gone yourself, because “the update should have removed it” is not a check.
On OPSSHIELD’s side, the attacker also copied about 7,900 billing-portal account records: names, companies, postal addresses, email addresses and hashed passwords. Those passwords were stored as bcrypt hashes, not plain text, which slows cracking down a lot — but OPSSHIELD still asks every customer to reset their billing portal password as a precaution. If you have a cPGuard license, that means you.
One honest caveat: the server and account counts come from OPSSHIELD and have not been independently audited, and the company’s investigation into the attack path is still open. The numbers could move. What has not moved is the advice — every independent write-up of the incident, including BleepingComputer’s coverage on 3 November, lands in the same place: treat an affected server as compromised even after the backdoor is removed.
Was Your VPS Affected? How to Check in 5 Minutes
Not every cPGuard user was hit. Impact depends on two things: whether your server had cPGuard installed, and whether it was online and checked for updates during the 02:08–03:20 UTC window on 2 November. Here is how to find out.
Step 1: check whether cPGuard is installed at all. Log in to your VPS over SSH and run:
which cpguard
ls /usr/local/cpguard 2>/dev/null
rpm -qa 2>/dev/null | grep -i cpguard; dpkg -l 2>/dev/null | grep -i cpguard
If none of those return anything, cPGuard is not on your server and you can stop here — the breach does not touch you. If you are on a managed VPS, you may not have root access to check; skip to the “managed VPS” section below and ask your host instead.
Step 2: check which version ran during the incident window. The fake update shipped as cPGuard versions 5.88.12 through 5.88.19. Check your installed version and your update history:
cpguard --version
grep -i cpguard /var/log/cron* /var/log/messages 2>/dev/null | grep -i "Nov 2" | head -20
ls -la /var/tmp/.cpg91511_sh /var/tmp/.i93677/ /etc/cron.d/i93677 2>&1
Step 3: read the result honestly. If your version is 5.88.12–5.88.19, or if any of those three backdoor paths exist on disk, your server ran the fake update — treat it as compromised and work through every step below. If your version is 5.88.26 or later and none of the backdoor files exist, you probably updated cleanly, but if your server was online during the window and you cannot prove which version it ran at 02:08 UTC, rotate the credentials anyway. The cost of rotating a password is five minutes; the cost of guessing wrong is a root-level compromise you discover months later.
Also check your billing portal account. If you ever bought cPGuard directly from OPSSHIELD, your name, email and hashed password were in the copied batch of about 7,900 records. Reset that password and turn on two-factor login whether or not your server was affected.

Step 1: Update cPGuard and Remove the Backdoor
Do this first, before rotating anything. There is no point changing passwords while a hidden root shell is still sitting on the box.
Update cPGuard to the latest stable release — as of 14 November that is version 5.91.01, with 5.90.00 released on 13 November. Do not stop at the first patched version you read about in early reporting; the vendor has shipped several releases since, and the latest one includes license-binding, update-script and permission changes on top of the backdoor removal. Update through cPGuard’s own updater or your control panel, then confirm the version:
cpguard --version
# Expect 5.91.01 or later — if it still shows 5.88.x, the update did not take
Then verify the backdoor is actually gone. The clean update is supposed to remove the hidden shell and the scheduled task, but verify it with your own eyes:
ls -la /var/tmp/.cpg91511_sh /var/tmp/.i93677/ /etc/cron.d/i93677 2>&1
# All three should report "No such file or directory"
# If any still exist, delete them manually:
rm -rf /var/tmp/.i93677/ /var/tmp/.cpg91511_sh
rm -f /etc/cron.d/i93677
While you are in there, check for the other indicators OPSSHIELD published. Look for probe files in your web root (/i90248_probe.txt, /i90573_probe.txt, /i90573_sh.php, /i90573_x.php — the attacker’s filenames follow an i9xxxx pattern), any *_rx.php endpoint you do not recognise, and connections to the attacker IP 94.183.188.91 in your logs:
grep -r "94.183.188.91" /var/log/ 2>/dev/null | head
find /home/*/public_html /var/www -maxdepth 2 -name "i9*.php" -o -maxdepth 2 -name "i9*.txt" 2>/dev/null
grep -r "_rx.php" /var/log/apache2/access.log /var/log/httpd/access_log 2>/dev/null | head
If you find probe files or web shells, that is evidence the attacker went further than the fake update on your server — jump to the rebuild section at the end of this guide instead of continuing with rotation alone.
Step 2: Rotate Root Passwords and SSH Keys
This is the step most people skip, and it is the one that matters most. The fake update copied /etc/shadow and your root SSH private keys. Updating cPGuard does nothing about that. Until you rotate them, the attacker holds credentials that work.
Change the root password and every local user password:
passwd root
# Then every other user with shell or sudo access:
passwd username
Use a fresh, long password for each — a password manager is the sane way to keep track. If you reuse the same root password across servers (be honest with yourself), change it on all of them, because the attacker now has the hash for every box that ran the fake update.
Generate new root SSH keys and kill the old ones. This is the part people get wrong: generating a new key is not enough — you also have to remove the old public key from every authorized_keys file that trusted it, or the stolen private key still logs in:
# On your own computer, generate a fresh key pair
ssh-keygen -t ed25519 -C "root@your-server-nov2026"
# On the server, replace the old key (do NOT just append)
cat ~/.ssh/id_ed25519.pub > /root/.ssh/authorized_keys
# Review authorized_keys for anything you do not recognise
cat /root/.ssh/authorized_keys
# And check every user's authorized_keys too
find /home -name "authorized_keys" -exec cat {} \;
Test the new key in a second terminal window before you close your current session — if the new key does not work and you have already removed the old one, you have locked yourself out. Once the new key works, the old private key is dead everywhere it was trusted. Also review known_hosts and any deploy keys, CI keys or backup scripts that used the old keypair; each one needs the new public key or it breaks silently later.

Step 3: Change the MySQL Root Password
The fake update read /root/.my.cnf, which usually holds the MySQL root password in plain text. If your VPS runs WordPress, a shop, or any app with a database, that password is now in the attacker’s hands — and with it, every database on the box.
# Log in to MySQL as root with the current password
mysql -u root -p
# Inside MySQL, set a new password:
ALTER USER 'root'@'localhost' IDENTIFIED BY 'Your-New-Strong-Password-Here';
FLUSH PRIVILEGES;
exit
Then update /root/.my.cnf with the new password so your own scripts and control panel keep working, and do the same for any application config files that store the MySQL root password. While you are there, check for database users you do not recognise:
mysql -u root -p -e "SELECT user, host FROM mysql.user;"
An unfamiliar user with broad privileges is a red flag — the fake update was a grab-and-go script, not a persistent implant, but if anything else got in through the stolen credentials afterwards, a rogue database user is exactly where it would hide. Remove anything you cannot explain, and change the passwords of your application database users too if they were stored anywhere root could read — which, on a typical cPanel box, is most places.
Step 4: Rotate Secrets From Shell History and Cloud Keys
Root shell history is a goldmine and most admins forget it exists. Every password, API token or secret you ever typed into a terminal — mysql -pSecretPassword, a curl call with a bearer token, an AWS key passed on a command line — is sitting in /root/.bash_history, and the fake update copied it.
Read your own history and make a list of everything sensitive in it:
grep -iE "pass|token|key|secret|aws_|api" /root/.bash_history | head -50
# Also check other users' history and root's history on every server that shares keys
cat /root/.bash_history | wc -l
Then rotate each one: API keys at the provider’s dashboard, tokens in your apps, and any password that appears in the history. After rotating, clear the history file so the old values stop sitting around in plain text:
history -c && history -w
# And stop the habit that created the problem:
# use config files or environment variables instead of typing secrets on the command line
Check /root/.s3cfg too. Only a small number of servers had cloud storage keys copied, but if yours is one of them, the attacker can read your backups and object storage. Rotate those keys at your storage provider and check the access logs for downloads you did not make — a backup bucket is the first place an attacker looks for database dumps.
Step 5: Check Every Server That Trusted Your Keys
A stolen SSH key does not stay on one server. If your compromised VPS had SSH access to other machines — a second VPS, a backup server, a staging box, a Git server — the attacker can walk sideways into all of them using the keys they copied.
Map the blast radius before you declare the job done:
# Which servers does this box reach with keys?
cat /root/.ssh/known_hosts | cut -d' ' -f1 | sort -u
grep -r "ssh " /root/.bash_history | grep -oE "[a-z0-9.-]+\.[a-z]{2,}" | sort -u | head -20
# Which servers trust THIS server's key? Check authorized_keys everywhere you manage:
grep -l "your-old-key-comment" /root/.ssh/authorized_keys /home/*/.ssh/authorized_keys 2>/dev/null
Every server on that list gets the same treatment: new keys, new passwords, and a check for the backdoor files. This is tedious and it is the step most people rush — but a single forgotten staging server with the old key still in authorized_keys hands the attacker a way back in after you have done everything else right. If you manage client servers, tell your clients. They would rather hear it from you now than discover it from a breach notice later.
On a Managed VPS? Ask Your Host These 5 Questions
If your host manages the server, you probably cannot run the checks above — and the responsibility sits with them. That does not mean you can ignore it. Your sites, your databases and your customers’ data live on that box. Send your host a ticket and ask, in writing:
- Do you use cPGuard on my server? Many managed hosts install it alongside cPanel without mentioning it. You need a yes or no, not a guess.
- Did an affected version (5.88.12–5.88.19) run on my server during 02:08–03:20 UTC on 2 November? They have the update logs; ask for the version history, not a reassurance.
- What version is running now, and were the backdoor files confirmed removed? The answer should be 5.91.01 or later, with the three artifact paths checked.
- Which credentials were rotated, and when? Root passwords, SSH keys, MySQL root password — if they say “we updated cPGuard” and nothing about rotation, the job is half done.
- Was my server rebuilt, or was rotation considered enough? And if you host client sites, ask what they told those clients.
Get the answers in the ticket, not on a phone call. If your host cannot answer question 2 — if they do not know which version ran during the window — treat the server as affected and ask them to rotate everything. A host that cannot prove you were safe has not proven you were safe. If the answers are vague or slow, that is useful information about your host too, and it might be the nudge you need to look at the cheapest VPS plans that do not cut corners on security when your term is up.
When to Rebuild the Server From Scratch
OPSSHIELD’s routine advice is update plus rotation — not a full rebuild. But the vendor itself calls that assessment provisional, and says to rebuild from a clean backup taken before 1 November where integrity cannot be established or additional compromise is found. Here is how to decide which camp you are in.
Rotate and move on if: your server ran the fake update, you found the known backdoor files (and removed them), nothing else looks wrong, and you complete every rotation step above. That covers most affected servers — the fake update was a short collection script, not a rootkit that buries itself in the kernel.
Rebuild from a clean backup if any of these are true:
- You found probe files, web shells or unfamiliar files in your web root
- You found logins from the attacker IP (94.183.188.91) or unfamiliar sessions around 2 November
- You found a database user, cron job, systemd service or SSH key you cannot explain
- OPSSHIELD told you the attacker logged in to your server by hand
- You cannot account for what happened on the server between 2 November and today
A rebuild means a fresh OS install, restore of data from a backup dated before 1 November, and re-applying every rotation step on the new box. It is a few hours of work and it is the only way to be sure nothing else is hiding. If you are unsure which camp you fall in, rebuild — the cost of a rebuild is a quiet afternoon, and the cost of a missed backdoor is your server sending spam, mining crypto or serving malware while your name is on the account.
Why Supply-Chain Attacks Are the New Normal for VPS Owners
The cPGuard breach is not an isolated weird event. It sits in a pattern that has been accelerating all year. In October, a critical PHP flaw (PHP2Shell) needed emergency patching on web servers everywhere — we covered the fix in our PHP2Shell patching guide. November’s Patch Tuesday brought four new RDP flaws, including a critical CredSSP issue — see our November 2026 RDP vulnerabilities roundup. The difference with cPGuard is that there was nothing to patch in advance. The vulnerability was not in code you run; it was in the trust relationship between your server and its vendor.
That is the lesson worth carrying forward. On a VPS, your security depends on a chain of vendors you never meet: the control panel, the malware scanner, the backup tool, the monitoring agent. Every one of them runs with high privileges and updates itself automatically — which is convenient until one of them is compromised. You cannot audit all of them, but you can reduce the blast radius:
- Separate your secrets. Do not reuse root passwords or SSH keys across servers. One compromised box should not open every box you own. Our VPS security checklist walks through key-only SSH login and disabling password authentication entirely.
- Keep backups off the server. A backup stored on the same box — or reachable with keys stored on the box — is not a backup against this kind of attack. Push backups to storage the server cannot read back without a separate credential.
- Watch your update logs. A quick weekly glance at what updated and when would have caught the 5.88.12–5.88.19 install on 2 November. Most admins never look.
- Prefer vendors that disclose fast. OPSSHIELD published a detailed timeline within a day, named the exact versions and indicators, and told customers to assume compromise. That is the standard to judge your other vendors by — a vendor that goes quiet during an incident is a vendor you should replace.
- Assume breach, then verify. The old mindset was “patch and move on”. The new one is “patch, rotate, verify, then move on”. It takes an extra hour and it is the only response that matches what actually happened.
The broader hosting world is feeling the same pressure. The RAM shortage that pushed VPS prices up through 2026 is squeezing small vendors hardest, and small vendors are exactly the ones attackers target — fewer engineers, thinner security budgets, and a customer base of small hosts who cannot absorb a breach. cPGuard’s price was its selling point for those hosts. That does not make it a bad product; it makes it a reminder that the cheapest tool in your stack still runs as root.
Frequently Asked Questions
Do I need to worry if my VPS runs cPanel but I never installed cPGuard?
Probably not, but check. cPGuard is often bundled by hosts or installed by a previous admin and forgotten. Run the checks in the “Was Your VPS Affected?” section — if nothing is installed, the breach does not touch you. If you are on managed hosting, ask your host directly; “we use cPanel” does not answer whether cPGuard is also on the box.
Is updating to cPGuard 5.91.01 enough on its own?
No. Updating removes the backdoor and the scheduled task, but the fake update already copied your password hashes, SSH keys and MySQL root password. Those stay valid until you rotate them. Update first, then rotate everything — in that order.
What if my server was offline during 02:08–03:20 UTC on 2 November?
Then it did not receive the fake update, and your server secrets were not copied. You should still reset your OPSSHIELD billing portal password, because the 7,900 copied billing records are a separate exposure from the server-side theft.
Should I uninstall cPGuard entirely?
That is a fair question and there is no single right answer. The vendor has revoked the stolen keys, rebuilt its infrastructure, introduced a privately signed update channel in 5.89.00, and shipped hardened releases through 5.91.01. If you keep it, update to the latest release and keep watching the vendor’s advisories. If you drop it, replace it with another scanner rather than running a public-facing server with no malware scanning at all — and check our new VPS setup checklist for the rest of the hardening steps that should go with it.
How do I know if the attacker used my stolen credentials after 2 November?
Check your auth logs for logins you do not recognise, especially around and after 2 November: grep "Accepted" /var/log/auth.log (or /var/log/secure on RHEL-based systems). Look for logins from unfamiliar IPs, logins at odd hours, and any new users or keys. Also check whether your server sent unusual outbound traffic — a spike in the days after 2 November is worth investigating even if the logs look clean.
Does this affect shared hosting customers?
Indirectly. If your shared host’s servers ran cPGuard, the host is responsible for the response, not you — but your site’s database credentials and files live on those servers. Ask your host whether they were affected and what they rotated. If they cannot answer, that tells you something about the host.
Where can I read the vendor’s full disclosure?
OPSSHIELD published the full timeline, the stolen-data list and the indicators of compromise on 3 November. Read it here: OPSSHIELD security incident disclosure.
The Bottom Line
The cPGuard breach is a reminder that on a VPS, the software you trust most is the software that can hurt you most. A malware scanner with root access, an automatic updater and a compromised vendor became a skeleton key for about 2,950 servers in 70 minutes — and the fix is not just an update, it is an hour of unglamorous rotation work: new passwords, new SSH keys, a new MySQL root password, and every secret your shell history ever recorded.
Do it today, while the details are fresh and the indicators are published. Check whether cPGuard is on your server, check the version, update to 5.91.01 or later, confirm the backdoor files are gone, and then rotate everything the fake update could have read. If you find anything you cannot explain, rebuild from a backup taken before 1 November. It is a boring afternoon of work — and it is the difference between a server you control and a server somebody else has the keys to.


