Data quality drives detection
A SIEM is only as good as the data it receives. Common problems include unparsed JSON blocks, missing or local timestamps, inconsistent field names and large volumes of debug output that cost storage and attention.
Proven parsing techniques
- Normalise timestamps: convert all event times to ISO 8601 / UTC at ingestion. Local agent clocks drift and make correlation harder.
- One schema: use the same field names across all sources, such as
src_ipanddest_ipfor firewall, EDR and proxy. Standards like ECS or OCSF help. - Enrich with context: link events to asset information (criticality, owner, vulnerabilities) so analysts can prioritise immediately.
- Multi-line logs: merge stack traces and multi-line events at the agent, not in the SIEM.
Reducing noise safely
Dropping logs blindly is risky. Instead:
- Filter high-volume benign events at host level, for example with a granular Sysmon configuration.
- Introduce hot-warm-cold storage so older audit logs stay available at low cost.
- Align detection rules with MITRE ATT&CK and test them regularly with controlled simulations.
Conclusion
A clean pipeline saves licence costs, speeds up investigations and reduces alert fatigue. Whether Wazuh, Elastic or Splunk, we help you build and tune it.
