Detection Engineering

Log Pipeline & Normalisation

Raw log data from different sources cannot be correlated without normalisation.

What it is

A SIEM ingesting logs from Windows, Linux, cloud platforms, firewalls, and identity providers receives logs in dozens of different formats. Without normalisation — mapping source-specific field names to a common schema — correlation rules cannot operate across sources. A login event from Azure AD and a login event from a VPN appliance have different field names, different timestamp formats, and different event structures. Detection rules that require correlation across both will not fire without a normalisation layer.

Attack techniques

  • Inconsistent field naming across log sources
  • Timestamp timezone mismatches breaking correlation
  • Missing enrichment (hostname to IP resolution)
  • Raw log ingest without schema mapping
  • Parser failures silently dropping log lines

Business impact

Correlation rules fail silently when log normalisation is broken. The SIEM appears to be working but cross-source detections never fire.