A Linux server is rarely compromised by an advanced exploit alone. It is compromised by an old package, a password-only SSH login, an application running as root, or a service that was opened to the internet for a one-off task and never closed. Hardening is the discipline of removing those easy paths before someone else finds them.
This checklist is written for a small team that manages its own servers, whether on a VPS, a colocation rack or a cloud account. Work through it in order, record evidence for each item and schedule the work in a maintenance window. Do not treat it as a one-time audit. The checklist is most useful as a quarterly routine.
Prerequisites
- An accurate inventory: hostname, owner, purpose, operating system, exposure to the internet and data classification.
- Out-of-band access: console access through the provider, not only SSH, so a firewall mistake does not lock you out.
- A tested backup and a rollback plan: a snapshot is not a backup until you have restored from it.
- A change window and an approver: hardening changes can break applications, so agree the window and the rollback trigger.
- A baseline capture: open ports, installed packages, users, cron jobs and running services before you change anything.
Capture the baseline with commands such as ss -tulpn, systemctl list-unit-files --state=enabled, getent passwd and crontab -l, and store the output with the change record.
Assign every section to a named owner before you start. In a small team one person may own several sections, but the name and the review date should still be recorded.
1. Patch and package baseline
- Enable automatic security updates, or a weekly patch window with an owner.
- Remove packages and services the server does not need. Every extra package is extra attack surface.
- Confirm that package repositories are official and signed. Remove third-party repositories you no longer use.
- Track kernel and library versions, and reboot when a kernel update requires it.
- Keep an inventory of end-of-life software and a dated plan to replace it.
2. Accounts and authentication
- One named account per administrator. No shared root login.
- Use
sudowith specific permissions rather than unlimited root access for everyone. - Disable direct root login over SSH and require key-based authentication with a passphrase on the key.
- Remove or lock accounts for former staff and unused service accounts.
- Review accounts with UID 0 and any account with a valid shell.
- Set a strong password policy for local console access, even when SSH uses keys.
- Require MFA for the bastion host or the cloud console that controls the server.
3. SSH and remote access
- Set
PermitRootLogin no,PasswordAuthentication noandPubkeyAuthentication yes. - Restrict access with
AllowUsersorAllowGroupsinstead of allowing every account to connect. - Limit
MaxAuthTriesand idle timeout values. - Bind SSH to a private interface or require a VPN or bastion for administrative access.
- Treat tools such as Fail2ban as noise reduction, not as a substitute for key-only access.
- Revoke user keys when someone leaves or a laptop is lost, and rotate host keys only during a planned rebuild or a suspected compromise.
4. Services, ports and network exposure
- Default-deny inbound traffic with the host firewall, then allow only required ports.
- Close database, cache, admin panels and monitoring ports to the public internet. Bind them to localhost or a private network.
- Disable or remove unused services, including print, mail and file-sharing daemons.
- Document every public port with a business reason and an owner.
- Restrict outbound traffic where practical, especially from web servers and containers.
- Review cloud security groups and network ACLs at the same time as the host firewall.
5. Filesystem and permissions
- Set a restrictive default
umask, typically027for servers. - Mount
/tmp,/var/tmpand/dev/shmwithnoexec,nosuidandnodevwhere applications allow it. - Find and review world-writable files and directories.
- Review SUID and SGID binaries and remove unnecessary ones.
- Separate application data, logs and operating system files onto appropriate partitions.
- Use file integrity monitoring on critical binaries and configuration files.
6. Logging and monitoring
- Send authentication, sudo, cron and application logs to a central location the server cannot silently delete.
- Enable and tune audit rules for privilege changes, sensitive file access and unexpected process execution.
- Synchronise time with a reliable source so correlations work.
- Alert on repeated login failures, successful logins from new locations and changes to firewall rules.
- Define log retention based on business and regulatory needs, and test that logs are actually arriving.
7. Web and application layer
- Run each application under a dedicated, unprivileged user.
- Keep the runtime, framework, CMS and plugins patched, and remove unused plugins and themes.
- Disable directory listing and verbose error pages in production.
- Terminate TLS with a current configuration and renew certificates automatically.
- Restrict upload directories so they cannot execute code.
- Keep secrets out of the web root and out of version control.
8. Backups and recovery
- Follow the 3-2-1 pattern: three copies, two media types, one off-site or offline.
- Encrypt backups and restrict who can delete them.
- Test a full restore, not just a file-level restore, at least quarterly.
- Record the recovery time and compare it with the business requirement.
- Keep a restore runbook that a second person can follow.
Evidence and verification
For each section, keep a short evidence pack: the command output, a screenshot of the relevant configuration, the date, the person who checked it and any exception with a review date. A checklist without evidence becomes a document that everyone signs and nobody verifies.
Common mistakes
- Hardening a server without a rollback path and breaking production.
- Disabling password login before confirming that key access works for every admin.
- Opening a database port "temporarily" for a report and forgetting to close it.
- Relying on a single security tool while leaving default accounts active.
- Treating a vulnerability scan as proof that the server is hardened.
- Skipping the restore test because the backup job reports success.
Metrics to track
- Percentage of servers with key-only SSH and no direct root login.
- Median days to apply critical patches.
- Percentage of hosts sending logs to the central collector.
- Number of public ports without a documented owner.
- Backup restore test pass rate and measured restore time.
- Number of stale accounts removed in the last quarter.
Where this fits in our work
If you need a hardened baseline without adding another full-time role, start with Managed Technology & Cybersecurity. The pricing page explains how we scope server and infrastructure work. For the web layer, read the malware prevention checklist for business websites.
Sources
- Center for Internet Security, CIS Benchmarks, accessed 17 November 2026.
- NIST SP 800-123, Guide to General Server Security, accessed 17 November 2026.
- OpenBSD, sshd_config manual page, accessed 17 November 2026.
- NIST SP 800-63B, Digital Identity Guidelines: Authentication and Lifecycle Management, accessed 17 November 2026.
- OWASP Application Security Verification Standard, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.