The Problem and Why It Matters
In August 2025, a malicious actor breached the systems of Miljödata, an IT service provider, exposing data on the darknet. This incident affected 2.2 million individuals, including customers like Sweden's municipalities, regions, government agencies, and private companies. The compromised data included personal identity numbers, contact details, and sensitive records on sick leave, rehabilitation, and school incidents.
The Swedish supervisory authority (IMY) identified two major gaps: inadequate checks during software installation and the absence of automated real-time monitoring to detect intrusions. IMY fined Miljödata SEK 1,800,000 (approximately EUR 160,000) for violating Article 32(1) GDPR.
These issues are solvable. This guide will help you address them. As a Data Protection Officer (DPO), you won't build the system yourself, but you can ensure your security team meets a defined standard and provides evidence that your technical and organizational measures match the risk of your processing.
What You Need Before Starting
Gather these inputs to avoid generating useless noise:
- A processing inventory with sensitivity ratings. Article 32(1) ties security levels to risk. Identify systems holding special category data or national identifiers for tight detection coverage.
- An asset list covering every system that handles personal data, including processor-hosted environments. Check your Article 28 contract for logging access and incident notification duties.
- A log source map. Include identity provider, endpoints, servers, databases, cloud control plane, email gateway, and network egress.
- A SIEM or log analytics platform with alerting. Options include Microsoft Sentinel, Elastic Security, Splunk, or Wazuh for a budget-friendly start. Ensure alerts reach someone on call.
- An owner and an escalation path. Designate who triages alerts and informs the DPO. Article 33 requires notifying the supervisory authority within 72 hours of a personal data breach, so keep the alert path short.
Step-by-Step Implementation
Step 1: Centralize Important Logs
Forward authentication, privilege, and data access events to a central location. Focus on:
- Identity provider sign-in and audit logs (failed logins, MFA changes, new admin roles).
- Endpoint detection telemetry from servers and workstations.
- Database audit logs for systems with sensitive records. Use
pgauditfor PostgreSQL, SQL Server Audit for SQL Server, and the appropriate audit plugin for MySQL. - Cloud control plane logs (AWS CloudTrail, Azure Activity Log, or GCP Cloud Audit Logs).
- Firewall and proxy logs showing outbound traffic.
Set retention long enough to investigate slow intrusions, and restrict who can modify or delete logs. Logs that can be wiped by an attacker are useless.
Step 2: Write Detections for Key Behaviors
Avoid generic alerts. Focus on behaviors that precede mass data exfiltration:
- A service account or user reading unusually large volumes from a sensitive table.
- Bulk exports or large outbound transfers to unfamiliar destinations, especially outside business hours.
- New admin accounts, new MFA devices, or privilege escalation shortly before data access.
- Logins from impossible travel locations or new geographies.
- Security tools disabled or log forwarding stopped. A silent log source is an alert in itself.
Map each rule to a log source from Step 1 and assign an owner for tuning.
Step 3: Route Alerts to a Human Automatically
Connect the SIEM to your paging or ticketing tool so high-severity alerts reach the on-call responder without manual checking. Define severity tiers and response targets, and document them. Automated containment, like disabling a compromised account or isolating an endpoint, reduces the time between detection and damage. Keep containment actions simple and reversible.
Step 4: Implement a Software Gate
IMY criticized inadequate checks during software installation. Create a mandatory intake process:
- Request and risk triage. Submit any new software or updates with the personal data it will touch.
- Pre-deployment checks. Run a vulnerability scan, review dependencies (using OWASP Dependency-Check or Trivy), and ensure default credentials and open ports are removed.
- Configuration baseline. Compare against a hardened baseline like a CIS Benchmark.
- Approval record. Document who approved, what was tested, and any residual risk accepted.
- Post-install verification. Ensure the new system sends logs to the SIEM before going live.
Make this gate a condition for production access. Apply the same process to processor-supplied software and updates.
Step 5: Document Measures Against Article 32(1)
Create a record linking each control to the risk it addresses. Article 32(1) expects measures appropriate to the state of the art, costs, and the nature, scope, and risk of processing. A simple mapping table (risk, control, owner, evidence location) suffices and is the first thing a supervisory authority will request.
Validation: How to Verify It Works
Don't assume detections work. Test them.
- Run a log coverage test. Generate a known event for each critical system and confirm it appears in the SIEM within your target window.
- Simulate behaviors. Use a framework like Atomic Red Team to trigger detections: a bulk read, a new admin account, a disabled agent. Record which alerts fired, how fast, and who responded.
- Test the alert path end to end. Page the on-call responder at an inconvenient hour and measure the time to acknowledgment.
- Audit the software gate. Sample recent installations and confirm each has a pre-deployment record and post-install log forwarding.
- Exercise the Article 33 workflow. Conduct a tabletop exercise where a simulated detection must reach the DPO and produce a notification decision within 72 hours.
Keep the results. Dated test evidence shows your measures work, not just that they exist. Article 32(1)(d) requires regular testing and evaluation of effectiveness.
Maintenance and Ongoing Tasks
Monitoring requires regular attention.
- Weekly: Review alert volume and false positives; adjust or remove rules that waste time.
- Monthly: Check for silent log sources and ensure new systems are onboarded to the SIEM.
- Quarterly: Repeat a detection simulation and review access to log data.
- After significant changes: Re-run the software gate and update the Article 32 mapping record.
- Annually or after an incident: Reassess risk ratings against your processing inventory and test the breach notification workflow again.
As DPO, regularly ask: what sensitive systems lack detection coverage, and when was the last alert tested from trigger to human response? If you can't answer both with dated evidence, you're at risk of the same exposure IMY penalized.




