Hostinger’s LiteSpeed Breach: What to Rotate in WordPress

Shared hosting rests on one promise: lots of strangers’ websites share a server, and none of them can touch yours. On 9 November 2026 Hostinger published an incident report describing a time that promise broke. An attacker got into one customer’s site, broke out of it through a previously unknown hole in LiteSpeed Web Server, got root on the whole machine and “ran code or read credentials on about 400 accounts.”
Last updated 18 November 2026. We read Hostinger’s report, LiteSpeed’s changelog and the coverage we could find, and ran the WP-CLI commands below on our own test site (WordPress 7.0.2, WP-CLI 2.12.0). We didn’t test the exploit and we aren’t Hostinger customers.
This post is about the bit the report leaves out: Hostinger restored the affected sites from backups, but a backup brings back your files. It doesn’t un-steal a password. If your host ever tells you something like this happened on your server, here’s what we’d change, in order.
What happened, in plain words
Here’s the timeline as Hostinger tells it, plus LiteSpeed’s own changelog for the fix:
| Date (2026) | What happened | Source |
|---|---|---|
| 5 November | The attacker “first entered” through “a vulnerable content management system on a customer website”. | Hostinger |
| 5 to 8 November | “It took three days for the attacker to find and exploit a new vulnerability in LiteSpeed Web Server.” The bug let code from one site bypass CloudLinux CageFS and led to root on the server. | Hostinger |
| 8 or 9 November | Detection. The report’s body says monitoring “flagged suspicious activity” on 8 November; its opening line says “On November 9, 2026, we detected unauthorized access”. | Hostinger |
| 9 November, 12:26 UTC | Incident declared. From 13:30 UTC, accounts moved to other servers and affected sites “restored from backups taken before the attack”. | Hostinger |
| 10 November | LiteSpeed Web Server 6.3.7 Build 9 released: “[Security] Enhance Apache configuration handling to prevent unauthorized file access beyond the document root and running scripts without suEXEC (CageFS escape).” | LiteSpeed changelog |
| 10 November, 16:03 UTC | Build 9 installed on all Hostinger shared hosting servers. | Hostinger |
About that 8-versus-9 November gap: it looks like “monitoring flagged something” on the 8th and “we confirmed it” on the 9th, but the report doesn’t say so, so we’re just pointing it out.
The scale: one server out of “more than 5,000”. Hostinger says it found “no evidence that the attacker compromised our internal management systems, our backup system, or the systems that handle logins to Hostinger accounts, and no evidence that any backups were decrypted.” The original server is offline and the investigation is ongoing.
Two more things we checked. LiteSpeed’s changelog entry for Build 9 lists no CVE number, and we didn’t find one anywhere else. We also found no public statement from CloudLinux about this incident as of 18 November. And this isn’t Hostinger’s first one this autumn: its September report describes a different LiteSpeed zero-day on a Brazil server, where the attacker “deployed a webshell on 430 accounts”.
Why it matters even if you’re not on Hostinger
Nothing here is specific to Hostinger. LiteSpeed Web Server is the software that answers web requests, used in place of Apache or Nginx, and plenty of shared hosts advertise it. CloudLinux is an operating system built for shared hosting, and CageFS is its wall that keeps accounts on the same server apart: each account sees its own files and its own processes, and nobody else’s.
That wall is the reason shared hosting is safe enough to use at all. Here, a bug in the web server let code from one site walk around it. Once the attacker was root, every account on the box was readable.
The uncomfortable part is that your own site did nothing wrong. The way in was someone else’s outdated CMS. Your plugins can be up to date, your passwords strong, your firewall plugin on, and none of it helps when the hole is in the server underneath WordPress. (That’s also the limit of the firewall-style protection we covered in our post on virtual patching: it filters requests to your site, not a neighbour’s.)
Restored from backup is not the same as safe
Hostinger did the right things on the file side. It moved accounts, restored sites from pre-attack backups and closed the hole. If the restore is complete, any backdoor dropped into your files should be gone.
But look at what the report says was taken: “database credentials and access tokens.” Those are secrets, and a stolen secret keeps working until someone changes it. Restoring last week’s wp-config.php restores last week’s database password, which is exactly the one the attacker read.
The public report gives customers no advice on this. It doesn’t mention database passwords, WordPress salts or API keys. Hostinger says it “emailed affected customers individually”, and we haven’t seen those emails, so they may well say more. If you got one and it doesn’t tell you to rotate anything, do it anyway.
What we’d rotate, in order
This is for a WordPress site on any shared host after an incident like this one. If your host confirms your account wasn’t touched, you mostly don’t need any of it. It’s still a decent once-a-year habit.
1. Ask your host whether your account was affected
Do this first, because the answer decides how far down this list you go. Open a ticket and ask three plain questions: was my account on the affected server, was any code run or any file read in my account, and what date was my restore point? That last one matters: if your site was rolled back, check that recent posts, orders or form entries didn’t vanish with it.
2. Change the database password
This is the credential the report names first. Anyone holding it, plus a way to reach your database, can read every user, order and setting, or quietly add an admin.
- In your hosting panel, find the database user your site uses (in cPanel: MySQL Databases; other panels have a Databases page). Not sure which one? It’s
DB_USERinwp-config.php. - Set a new long random password.
- Straight away, put the same password into
wp-config.php:define( 'DB_PASSWORD', 'new-password-here' );
Between steps 2 and 3 your site shows “Error establishing a database connection”, so have both windows open. While you’re in the panel, check any “Remote MySQL” allow list for hosts you didn’t add.
3. Replace API keys and tokens your site stores
“Access tokens” is vague, and Hostinger doesn’t say which ones. So assume any key your site keeps was readable. Most plugins store them in the wp_options table in plain text, and some live in wp-config.php. The usual list:
- SMTP passwords or email API keys (SendGrid, Mailgun, Amazon SES and so on)
- Payment gateway secret keys and webhook secrets (Stripe, PayPal)
- AI provider keys. We looked at where WordPress keeps those in where WordPress stores your AI API key.
- Backup destinations (Dropbox, Google Drive, S3 keys inside your backup plugin)
- Cloudflare or CDN API tokens, and any “connect your site” tokens for remote management tools
The right move is to revoke the old key at the provider and make a new one there. Just pasting a new key into WordPress leaves the old one alive.
4. Shuffle the WordPress salts
The eight AUTH_KEY, SECURE_AUTH_SALT and friends in wp-config.php are secret values WordPress uses to sign login cookies. Together with a copy of your database (which this attacker could reach through the stolen credentials), they’re enough to forge a login cookie. Changing them makes every existing login cookie invalid, so everyone (you included) gets logged out. Passwords keep working.
With WP-CLI it’s one command: wp config shuffle-salts. Without it, get fresh values from https://api.wordpress.org/secret-key/1.1/salt/ and paste them over the old block. A few plugins use these keys to encrypt their own saved settings, so be ready to reconnect one or two afterwards. One we’ve checked ourselves: the official AI plugin’s Key Encryption experiment builds its encryption key from LOGGED_IN_KEY and LOGGED_IN_SALT unless you’ve defined WP_SECRETS_KEY, so after a shuffle expect to paste your AI API key in again. Since that key may have been read too, a fresh one is what you want anyway.
5. Reset admin passwords and application passwords
Change the password for every administrator under Users > Profile. On the same screen, scroll to Application Passwords and revoke any you don’t recognise, or all of them. Those let apps and scripts log in without your main password, and they survive a password change. The WP-CLI version:
wp user application-password list admin
wp user application-password delete admin --all
Also change your SFTP/FTP passwords in the hosting panel. If you use SSH, check ~/.ssh/authorized_keys for keys that aren’t yours.
6. Look for anything left behind
A restore should clean the files, but a quick look costs five minutes. This is the same sweep we suggested after the UpdraftCentral backdoor wave:
wp user list --role=administrator --fields=ID,user_login,user_email,user_registered
wp plugin list --status=must-use
wp core verify-checksums
wp plugin verify-checksums --all
find wp-content -name '*.php' -newermt '2026-11-01'
What you’re after: admins you didn’t create, “must-use” plugins you didn’t add (they live in wp-content/mu-plugins, run on every page and don’t appear in the normal Plugins list), and PHP files changed recently that you didn’t change.
A warning from running these on our own test site: the checksum commands are noisy. verify-checksums flagged one core file it didn’t expect to find, one plugin “failed” because it writes its own CSS files into its folder, and premium plugins get skipped because WordPress.org has nothing to compare them with. So a failure means “go and look at this file”, not “you’ve been hacked”.
How to tell what server your host runs
You can get a hint, not a full answer. From a terminal:
curl -sI https://your-site.example | grep -i '^server'
If it prints server: LiteSpeed, LiteSpeed is answering. Inside WordPress, Tools > Site Health > Info > Server shows the web server too, and so does a phpinfo() page (delete it afterwards).
The limits: behind Cloudflare the header usually says cloudflare, and some hosts hide it on purpose. None of these show the build number, so you can’t confirm 6.3.7 Build 9 yourself. And CageFS is invisible from your side. For both, ask your host: “Are you on LiteSpeed 6.3.7 Build 9 or later, and do you use CageFS?” A good host answers that in one line.
Frequently asked questions
Do I need to do anything if I’m on Hostinger but wasn’t contacted?
Probably not. Hostinger says it emailed affected customers individually and the attacker reached about 400 accounts on one server out of more than 5,000. If you want certainty, ask support whether your account was on the affected Phoenix server. Rotating your database password and salts is cheap insurance either way.
Is the LiteSpeed hole fixed?
LiteSpeed released 6.3.7 Build 9 on 10 November 2026 with a security fix its changelog labels a CageFS escape, and Hostinger says it installed that on all its shared servers the same day. Other hosts patch on their own schedule, so ask yours which build it runs.
Will changing WordPress salts break my site?
No. It logs everyone out, so all users have to sign in again, but passwords keep working. A few plugins that encrypt their settings with these keys may need to be reconnected.