In Part 1, we covered the application stack. In Part 2, we focused on the data layer. In Part 3, we addressed transactional email. This fourth installment closes the series with the question that actually matters when everything else fails:
Can the journal be fully and quickly recovered after a real incident?
If the answer is no, then you are not running OJS in production. You are running it in hope mode.
Why Backup and Security Are the Same Problem
Many institutions treat these as separate concerns:
- security to prevent intrusions
- backup to recover files
In OJS, that separation is artificial. A security incident that compromises the server will also compromise the backups if they live on the same host. And a backup that has never been verified is, in practice, no backup at all.
The real question is not whether the site is protected. The real question is whether the entire editorial record survives:
- ransomware or malicious deletion
- database corruption
- human error in production
- physical disk or VPS failure
- compromised SSH access
- failed OJS or plugin updates
If you cannot answer that with a verifiable restoration procedure, the issue is not only security. It is editorial continuity.
What an OJS Installation Actually Stores
OJS does not only store web pages. It stores the operational and academic record of a journal:
- submitted manuscripts
- supplementary files and proofs
- peer review rounds and comments
- editorial decisions
- users, roles, and permissions
- indexing metadata
- workflow notifications and traceability
Losing this data is not simply a website outage. It is an institutional loss with editorial, academic, and reputational consequences.
The Failure Scenario Most Journals Underestimate
The most common failure case is not a movie-style hacking story. It is something much simpler:
- A publicly exposed server receives unauthorized access through weak credentials, outdated software, or a vulnerable plugin.
- The attacker encrypts or deletes files.
- Local dumps and backups were stored on the same server.
- The institution discovers it had copies, but no recovery.
This pattern explains why local backups and reactive security are not enough.
A serious OJS backup must satisfy three conditions simultaneously:
- it lives outside the primary server
- it can restore the system end to end
- it has a retention policy that survives accidental deletion or malicious encryption
Local Backup vs Real Editorial Continuity
| Criterion | ❌ Basic backup | ✅ OJS One standard |
|---|---|---|
| Location | Same server or same volume | Off-site independent storage |
| Coverage | Only SQL dump or only files | Database + /files + public + config.inc.php |
| Verification | Assumed to work | Periodic documented restore test |
| Retention | Undefined | Daily, weekly, and monthly retention |
| Real goal | Having backups | Recovering the full journal with defined RPO and RTO |
What a Complete OJS Backup Must Include
If any of these components is missing, you do not have a complete restoration:
- MariaDB database with transactional consistency.
/filesdirectory containing manuscripts, reviews, proofs, and publication assets.publicdirectory andconfig.inc.php, plus any critical local customizations, custom plugins, or server-side scripts specific to the installation.
The base OJS source code itself is usually not part of operational backups. It can be reinstalled from the corresponding release or from version control. What must be preserved are the local files and configuration elements that cannot be reconstructed automatically.
Active installations should also record:
- exact OJS and plugin versions
- expected backup size or checksum
- date of the last successful restore test
- off-site destination and retention policy
Off-Site Is Not Optional
Saving backups in another directory on the same VPS is not off-site. Storing them on another mounted volume of the same server is not off-site either.
Off-site means the copy lives on independent infrastructure: S3-compatible object storage or a separate SFTP destination with independent credentials.
And one clarification matters: VPS snapshots do not replace real backups. Snapshots can help with fast rollback after an operational mistake or a failed update, but they still depend on the same provider, the same environment, and the same retention model. In OJS, snapshots are complementary. Verified off-site backups are what actually reduce loss risk.
Defense in Depth: Minimum Security for OJS in Production
Backup reduces impact. Hardening reduces probability. Serious OJS operations require both.
1. Perimeter WAF
Cloudflare or an equivalent layer should filter automated traffic, aggressive scraping, and abnormal spikes before they reach the server.
2. Minimal Port Exposure
Only strictly necessary ports should be exposed. MariaDB, Redis, and administrative panels should not be publicly reachable from the Internet.
3. Separation of Sensitive Storage
Editorial files should not live inside the public web tree served by Nginx or Apache. In OJS, this means keeping document storage outside the webroot. This is classic hardening: it prevents direct exposure of manuscripts, reviews, temporary proofs, and other sensitive files when routing or web server rules are misconfigured.
4. SSH Public Key Authentication
No password-based SSH for privileged access. Use named users, public keys, and least privilege.
5. Proper TLS Configuration
OJS handles unpublished manuscripts, credentials, and peer review communication. Encryption in transit is not cosmetic.
6. Controlled Update Surface
The operating system, PHP, Nginx, MariaDB, OJS, and plugins should remain on supported versions. Every unnecessary plugin increases attack surface.
7. Operational Monitoring
Security is not something you operate only when the site goes down. You need alerts for uptime, disk usage, backup failures, abnormal activity, and server-side errors.
And in OJS there is one signal that should not be missing: slow query visibility in MariaDB. This is not just a performance tool. It provides operational visibility into editorial workflow degradation, table growth, costly filters, and queries that start breaking response times before they become incidents.
Recovery Metrics That Matter
RPO — Recovery Point Objective
Maximum acceptable data loss between the last valid backup and an incident. In production OJS, this should usually be less than 24 hours.
RTO — Recovery Time Objective
Maximum time required to restore the journal to an operational state. In a managed and documented environment, this should be measured in hours, not days.
Without restore tests, both numbers are fiction.
16-Point Production Checklist
This is the minimum a serious installation should be able to answer “yes” to:
- There is a daily MariaDB backup with transactional consistency.
- There is a daily backup of the
/filesdirectory. public,config.inc.php, and critical local customizations are backed up.- Backups are copied to an independent off-site destination.
- Retention is defined across daily, weekly, and monthly layers.
- The latest successful backup can be verified by date, size, and logs.
- There is a documented restore test at least monthly or once per operational quarter.
- The server sits behind a WAF or perimeter proxy with basic mitigation.
- Only the required public ports are exposed and sensitive storage is separated from the webroot.
- MariaDB is not exposed to the Internet.
- SSH access uses public keys and least-privilege users.
- TLS is enforced and insecure access redirects to HTTPS.
- The operating system and OJS remain on supported versions.
- Unnecessary plugins have been removed or disabled.
- Alerts exist for backup failure, downtime, disk saturation, and slow-query diagnostics.
- There is a clearly defined owner for recovery and escalation procedures.
If an installation fails several of these points, it does not have an isolated weakness. It has accumulated operational debt.
What Disqualifies a Production Setup
- Backups stored on the same server as production
- Reliance on VPS snapshots as if they were complete backups
- No restore testing
- Publicly exposed MariaDB
- Sensitive editorial storage inside the webroot
- Password-based SSH for privileged access
- Unreviewed or abandoned plugins in production
- No visibility into failed backups or operational degradation
Have your OJS backups ever been restored in a real test? The managed OJS plans include tested backups and monitoring, and if you need to harden an existing installation, the OJS cybersecurity service starts with an audit of your server.
FAQ
How often should an OJS journal be backed up?
At minimum, once per day. For journals with heavier editorial throughput or multi-journal installations, daily full backups should be complemented with snapshots or additional replication according to the target RPO.
Does a VPS snapshot count as a backup?
Not by itself. A snapshot is an additional fast-recovery layer, but it does not replace a consistent database dump, a /files copy, backup of public and config.inc.php, or independent off-site storage with restore testing.
Is exporting the database enough?
No. A SQL dump without /files does not recover manuscripts, attachments, or proofs. And without configuration, rebuilding the environment becomes slow and error-prone.
Which matters more: the WAF or the backup?
They do not compete. The WAF reduces the probability of an incident. The backup reduces the impact when the incident still happens. Serious OJS operations require both.
How do I know whether my backup really works?
Restore it into a separate environment and confirm that the journal opens, files exist, users can authenticate, and a historical submission retains its complete workflow traceability.
Conclusion
OJS is not just a CMS for scholarly publishing. It preserves editorial decisions, evidence of peer review, and institutional scholarly output. That is why, in production, backup and security are the same problem viewed from two different moments: before and after the incident.
The standard is not having copies. The standard is being able to restore the entire journal quickly, with integrity, and with a verified procedure.
Cybersecurity for OJS
Is your OJS patched and free of prior compromise?
We audit the installation, apply OJS and PHP patches with evidence, hunt for webshells and persistence, and respond if an incident has already happened.
- OJS-specific audit and hardening
- Patches with a tested restore
- Incident response
Cybersecurity for OJS
Is your OJS patched and free of prior compromise?
We audit the installation, apply OJS and PHP patches with evidence, hunt for webshells and persistence, and respond if an incident has already happened.
Review my OJS