On Monday 9 November 2026 Hostinger published an incident report about one of its shared hosting servers, which it calls Phoenix. The report is signed by Saulius Lazaravičius, its Chief Product Officer, and it is unusually specific. It is also the clearest real example we have seen this year of a risk every shared platform carries, our own included.
What happened, in Hostinger's account
- The way in. On 5 November an attacker got in "through a vulnerable content management system on a customer website". That was one customer's out-of-date CMS, not a flaw in the platform.
- The breakout. Three days later the attacker found and used a previously unknown vulnerability in LiteSpeed Web Server. Hostinger says it let code from one website bypass the isolation between accounts provided by CloudLinux CageFS, and eventually gave the attacker root on the server.
- The blast radius. The attacker "ran code or read credentials on about 400 accounts, including database credentials and access tokens". That was on one of the more than 5,000 servers Hostinger runs.
- Why it was hard to see. The attacker's code ran as the shared web server user, so in Hostinger's words its activity "blended in with normal server traffic".
The response is just as specific. Hostinger's own monitoring flagged the activity on 8 November, and the incident was declared at 12:26 UTC on 9 November. From 13:30 UTC affected accounts were moved to other servers and their sites restored from backups taken before the attack. The original server was taken offline and kept for forensics, and internal credentials were rotated. 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." It emailed affected customers individually, notified the competent authority, and is working with LiteSpeed on a fix.
That deserves credit. The breach was detected one day, declared the next, and published the same day with an entry point, a mechanism and a count of affected accounts, under an executive's name. A host that publishes this is doing better than one that does not. Everything below relies on details Hostinger chose to share.
The part that applies to everyone
The customer whose CMS was out of date was one account. The other accounts, about 400 in all, had done nothing wrong. Their data was read because of a neighbour's patching and a bug in a layer they all shared, and no customer could have audited either.
That is not unique to one host or one product. Any platform where many customers share something underneath, whether a web server, a kernel or a machine, has the same shape of risk. DeployBase is in that class too: every app runs in its own container, but the containers share a Linux kernel. We wrote about what that means when a kernel bug is announced in the patch is not yours to apply. That piece was about an advisory for a bug nobody had used. This one is a breach with a named entry point and a counted blast radius, and it answers the question that piece left open: what an attacker takes when the shared layer fails.
Hostinger's answer is specific: database credentials and access tokens. Those live in the same places in almost every web app, and that tells you what to do next.
The morning after: what to do if your host tells you
- Rotate the credentials your app holds. That means database passwords and every API token in your environment variables or config files: payments, email, storage, analytics, and any deploy or Git token. Create the new credential, deploy it, confirm the app works, then revoke the old one. In that order you avoid locking yourself out.
- Assume anything the web server could read was read. In this incident the attacker ran as the shared web server user. Treat
.envfiles,wp-config.php, framework config, and any backup or export file left inside the site folder as exposed, even if nothing looks touched. - Work out what each leaked token could reach, then check there. A database password reaches one database. A cloud storage key or a payment API key may reach far more. For each one, check that provider's own logs for the exposure window. Here that window is 5 to 9 November.
- Look for anything the attacker could have left behind. New admin users, unfamiliar files in upload folders, scheduled jobs you did not create, and changes to your application's code.
- Know your restore path before you need it. Hostinger restored affected sites from backups taken before the attack, which is the right move. Ask whether your host can do the same, how far back it can go, and whether you have a copy of your own that does not depend on the host at all.
On DeployBase, specifically
None of this stops a breakout, on our platform or any other. What a platform can do is make the morning after shorter. These are the parts of DeployBase that matter for the list above:
- Credentials set as environment variables live on each app's Environment page. For Node.js and Laravel apps, Save & Restart App puts a new value into effect straight away.
- Database backups can be created from the app's Access Details and downloaded, so you can keep a copy off the platform. Restoring one first saves a snapshot of the current database, so a restore can itself be undone.
- Git deployment (from Pro) builds what is in your repository. After an incident you can redeploy a known-good commit rather than trusting the files on disk.
Before anything happens
- Use a separate database user and password per app, so one leak opens one database.
- Give each API token the narrowest scope its provider allows.
- Keep backups and exports out of the folder your site serves.
- Keep your CMS, its plugins and its themes updated. You cannot patch your neighbours, but you can make sure the entry point is not you.
Source: Hostinger, "Unauthorized access to our Phoenix server: November 2026 incident report", Saulius Lazaravičius, published 9 November 2026, read in full 10 November 2026. Quotations are from that report; the description of the vulnerability as previously unknown, and its attribution to LiteSpeed Web Server and CloudLinux CageFS, are Hostinger's.