Shipping Is Only the Start: A Security Routine for Self-Hosted Apps
Your app works. Payments arrive, the dashboard loads, and the next feature is already half-built. Then a security advisory lands for something underneath it: the runtime, web server, container image or Linux distribution. Who checks whether production is affected while you are answering customers?
For an indie founder, that question matters more than collecting another security tool. Keeping an app secure is an operating habit, not a launch-day achievement. New vulnerabilities and changing access requirements need a repeatable response—even during a quiet month with no deployments.
Decide which responsibilities you actually own
A managed platform can take operating-system maintenance off your plate, depending on the service and plan. It does not automatically fix your authorization rules, exposed credentials, dependencies or account security. Read the provider’s responsibility boundary rather than assuming “hosted” means “handled.”
With a self-hosted Linux app, the list gets longer: host updates, firewall rules, SSH access, service configuration, backups and incident response. Containers package software; they do not transfer ownership of the host kernel or make old images safe. Choose self-hosting because its control is worth that work—not because the server invoice looks cheaper in isolation.
Make a one-page exposure inventory
Start with what someone can reach, not what your repository says you built. Record each production hostname, server, public service, runtime, database, storage bucket and administrative interface. Include staging systems with real credentials or customer data.
For each entry, note its owner, update source, support end date and recovery procedure. Mark whether it is internet-facing, restricted to a private network or accessible only through an administrative gateway. Verify that databases and dashboards are not accidentally public; close unnecessary ports and remove abandoned preview deployments.
Even a solo founder benefits from an owner column. “Me, with this checklist” is actionable. “Probably the hosting company” is not.
Turn patching into a verified outcome
Use supported Linux releases and runtimes, subscribe to their security notices, and schedule upgrades before support ends. Include application packages and container base images: updating the host does not rebuild the app’s image.
Ubuntu’s automatic-update guidance explains unattended security upgrades, permitted update sources and reboot options. Automation reduces repetitive work, but check failures and coverage. A third-party repository or pinned dependency may need a separate process.
After a change, verify the version actually serving requests, restart affected services when required, and confirm the intended kernel is running after any required reboot. Run a small smoke test: sign in, read and write a test record, and check the background worker. Keep a recovery path, but do not quietly leave a vulnerable rollback running indefinitely.
Protect the keys to the business
Enable MFA on hosting, source control, email and other administrative accounts; prefer phishing-resistant methods where supported. Keep recovery codes somewhere protected and accessible during an outage.
For SSH, use individual accounts and keys, limit administrative reachability, and grant elevated privileges only when needed. Test a second login and recovery access before tightening authentication so you do not lock yourself out. Remove former collaborators’ access promptly. Give the application a restricted service account rather than running everything as root.
Store secrets outside source control and public client bundles, separate production from development credentials, and restrict each token’s permissions. If a secret leaks, revoke or rotate it; deleting the visible copy is not enough. Apply the same restraint to coding agents: generating a feature does not require permanent access to every production credential.
Backups need a rehearsal
A successful backup job proves that a job ran, not that your business can recover. Keep a protected copy outside the server’s failure boundary, with access controls that prevent a compromised application from deleting every copy.
Restore into an isolated environment periodically. Check database contents, uploaded files and the configuration needed to start the app; measure how much data and time you would lose. Never let a test restore send real customer emails or charge cards. CISA’s ransomware guide recommends offline, encrypted backups and regular tests of their availability and integrity.
Prioritize your exposure, not the loudest headline
When an advisory arrives, match its affected versions and configuration to your inventory. Read the vendor’s remediation instructions. Use CISA’s Known Exploited Vulnerabilities catalog as an additional prioritization signal, not a complete list of everything dangerous.
An exposed, affected service with known exploitation deserves immediate attention; do not wait for Friday’s maintenance slot. If a fix cannot be applied promptly, consider vendor-recommended mitigations or temporarily removing the exposed feature. Record the decision and verification deadline. A quiet log does not establish that an unpatched service is safe.
Add visibility without outsourcing judgment
Collect authentication, application and system logs without unnecessarily storing secrets or sensitive payloads. Alert on meaningful changes: unexpected administrative access, suspicious process activity, failed backups and monitoring that goes silent. Test that an alert reaches someone who can act, and write down how to isolate a host and preserve evidence.
AI may increase the pressure. The UK NCSC’s assessment through 2027 expects AI to improve elements of intrusion operations, including exploitation of known vulnerabilities. That is an assessment, not proof that every attack is AI-driven or that fully autonomous attacks are inevitable.
Our experiment Unhaxor explores an AI security engineer for Linux monitoring and suspicious-activity alerts. Its public onboarding is opening in stages. Treat it as an experimental observation layer to evaluate, not a verified guarantee of detecting any particular exploit or a replacement for patching, backups and human response. Its stated scope is alerts, not automatic remediation.
Disclosure: MakerPad and Unhaxor share ownership.
Put a small routine on the calendar
Each week, review inventory changes, outstanding updates, access changes, backup results and unresolved alerts. Schedule restore rehearsals and larger upgrades separately. Keep critical vendor notices and security alerts able to interrupt that routine immediately.
Finish with a short record: what changed, what you verified, and what still has an owner and deadline. The goal is not a reassuring green dashboard. It is knowing who will notice, decide and act when your shipped app needs attention.