Skip to main content
Promotional banner ad for the Penetration Testing Report Kit
A VPN Move, a Year of Exposure: Six Migration QuestionsSecurity & Breach Notification
5 min readFor Privacy Officers

A VPN Move, a Year of Exposure: Six Migration Questions

A certificate authority moved an internal customer archive from behind a VPN to a publicly accessible network zone in February 2025. Nobody reviewed the resulting attack surface. The intrusion wasn't detected until June 2026, and a fix followed the next month. Some of the exposed records held passports, driver's licenses, and utility bills.

We've pulled together the questions that often arise in Slack channels and project steering meetings when a case like this circulates. The answers below are for those who need to sign off on migrations.


Q1: Our infrastructure team says a network move is "just plumbing." Do we really need privacy sign-off?

Yes, you do. Article 25 and Article 32 of the GDPR are relevant here. Article 25 requires data protection by design and by default, which applies when you change processing means, not just when you design them. Article 32 requires appropriate technical and organizational measures that account for risk, and a change in network exposure directly alters that risk profile.

In the DigiCert case, the system was moved from behind a VPN to a publicly accessible zone without a security review of the resulting attack surface. The data didn't change, but the exposure did. Add a trigger to your change management process: any change to network zone, authentication path, or public reachability of a system holding personal data requires a documented privacy and security review before going live.


Q2: What should that review actually check?

Keep it concise so it's actionable. At minimum:

  • Inventory what's reachable. Enumerate every endpoint, feature, and service the system exposes in its new location, not just the ones you intended to expose.
  • Compare old and new assumptions. The VPN acted as an access control. List what the old perimeter was protecting against.
  • Check dormant features. In this incident, an account registration feature became reachable from outside. The system was built for storing information, so nobody expected account creation.
  • Confirm authentication coverage. Every route to the data, including self-service and legacy ones, needs the same authentication requirements.
  • Test from outside. Run an external scan or penetration test against the migrated system before cutover, not after.
  • Re-run your DPIA screening. If the change raises risk to individuals, Article 35 may require an assessment or an update to an existing one.

Q3: We have MFA. Why does this keep biting people?

MFA is usually enforced per application or per identity provider, not per data store. In this case, the self-registered accounts weren't subject to multifactor authentication because the registration endpoint was never meant to be public. Those accounts had read access, which was enough to pull sensitive customer information from the archive.

Two lessons apply. First, "read-only" is not a mitigating label when the data is identity documents. Treat read access to sensitive records as a high-privilege capability. Second, audit the paths by which an account can come into existence, not just the accounts you know about. If a system can create identities, MFA enforcement and default permissions need to apply from the first login. Better still, ask whether the system should be able to create accounts at all.

MFA also has limits. It won't fix a data store that exposes records to any authenticated user. Pair it with least-privilege defaults, so a new account sees nothing until someone grants access deliberately.


Q4: How did this go undetected for more than a year?

The timeline tells you a lot. The exposure began in February 2025, the intrusion was detected in June 2026, and the fix came in July 2026. Even after the initial detection, the first conclusion was wrong: the team believed the unauthorized account couldn't reach data because it wasn't assigned to any application. The actual attack path was found a few days later, after a managed security program flagged the flaw.

Two monitoring gaps are visible. There was no alerting on blocked or anomalous access attempts against the archive, which the company later addressed by adding a SIEM alert for blocked access attempts. And the initial triage relied on an assumption about how permissions worked rather than verified evidence of what the account could actually do.

For your own program, build detection requirements into the migration checklist. Before cutover, confirm that the system's logs reach your monitoring stack and that someone has defined what abnormal looks like, such as new account creation on a system that shouldn't create accounts.


Q5: When the first triage says "no data accessed," how hard should we push back?

Push back firmly. Under Article 33, the controller must notify the supervisory authority without undue delay and, where feasible, within 72 hours of becoming aware of a personal data breach. Awareness is a judgment call, and a premature "no impact" conclusion can quietly burn your window.

Set an internal rule: a no-access finding needs evidence, not inference. Require the investigator to show what the account could reach, not what it was meant to reach. If the question is still open, treat it as open and keep the notification clock in view. Article 33(4) allows phased information where you can't provide everything at once, so you don't need to wait for a complete picture before engaging the supervisory authority.


Q6: The data included ID documents. Does that change our notification and communication duties?

It changes the risk assessment significantly. Article 34 requires communication to affected individuals without undue delay when a breach is likely to result in a high risk to their rights and freedoms. Identity verification documents are exactly the kind of data that enables fraud. As one security engineer commented publicly after the incident, a passport scan plus a utility bill is close to what some help desks accept to reset an account.

So your communication to individuals should be practical, not just formal. Tell them what was involved, and give them specific steps, such as adding a PIN or passphrase to their mobile carrier and bank accounts. Check the Article 34(2) content requirements: the nature of the breach, DPO or contact details, likely consequences, and the measures taken or proposed.

Also revisit retention. Ask why identity documents were still sitting in an internal archive. Article 5(1)(e) storage limitation and 5(1)(c) data minimization both apply. Verification documents often only need to be held long enough to complete verification.


Where to go for more

Start with the European Data Protection Board's guidelines on personal data breach notification, which cover awareness, risk assessment, and the 72-hour clock in detail. Read Article 25, Article 32, and Article 35 together as a set when you rewrite your change management gates. Then take one recent migration and run it back through the checklist in Q2. If you can't answer what the old perimeter was protecting, you've found your first gap.

GDPR full text

promotional banner for evidence for every CC6 control checklist

You Might Also Like