This Sysmon event ID reference gives detection engineers a practical validation checklist for Windows telemetry. Use it to confirm that expected events are generated, fields survive collection, and Sigma, Splunk, Microsoft Sentinel KQL, and Elastic detections still work after configuration or agent changes.
Overview
Sysmon is most useful when treated as a detection data source rather than a complete security control. An event can be enabled on the endpoint but missing from the SIEM, renamed during normalization, excluded by a filter, or stripped of fields during ingestion. A reliable validation process therefore checks the full path: benign test action, Sysmon configuration, Windows event channel, collector, parser, normalized fields, query, and alert.
The event IDs below are common starting points for a Windows telemetry review. Availability depends on the Sysmon version, operating system, configuration, and collector. Event IDs that are valuable for one investigation may create substantial volume in another environment, so enable and tune them deliberately in a lab before broad deployment.
| Investigation area | Sysmon events | Fields to validate | Safe test action |
|---|---|---|---|
| Process creation | 1 | UtcTime, Image, CommandLine, ParentImage, ParentCommandLine, User, IntegrityLevel, Hashes, ProcessId, ParentProcessId | Launch a signed utility with a distinctive, harmless argument. |
| Network connections | 3 | Image, ProcessId, Protocol, Initiated, SourceIp, SourcePort, DestinationIp, DestinationPort | Use a lab client to connect to an approved internal test listener. |
| Image loads | 7 | ImageLoaded, Image, ProcessId, Signed, SignatureStatus, Hashes | Open a normal application and record selected library loads. |
| File activity | 11, 15, 23, 26 | TargetFilename, Image, User, Hashes, IsExecutable, Archived | Create and delete a test file in a disposable directory. |
| Registry activity | 12, 13, 14 | EventType, TargetObject, Details, Image, User | Create, set, and remove a lab-only registry value. |
| DNS queries | 22 | QueryName, QueryStatus, QueryResults, Image, ProcessId, User | Resolve an approved internal or documentation domain. |
| Driver activity | 6 | ImageLoaded, Hashes, Signed, SignatureStatus, User | Install or load only an approved test driver in an isolated host. |
| Process access | 10 | SourceImage, SourceProcessId, TargetImage, TargetProcessId, GrantedAccess, CallTrace | Use a signed diagnostic tool against a process selected for testing. |
Other useful events include 2 for file creation-time changes, 5 for process termination, 8 for remote threads, 16 for Sysmon configuration changes, 17 and 18 for named pipes, 19 through 21 for WMI activity, and 25 for process tampering. Events 27 through 29 can support blocking or executable-file testing where the relevant Sysmon features and configuration are present.
Checklist by scenario
Process and PowerShell investigations
- Confirm Event ID 1 is generated for a benign command and that both
ImageandCommandLineare present. - Check parent-child context. Validate
ParentImage,ParentCommandLine, and process IDs rather than relying only on a process name. - For PowerShell detection, compare Sysmon process data with PowerShell operational logs and script-block logging. Sysmon can show the launcher and command line, but it is not a substitute for PowerShell-specific telemetry.
- Test a Sigma rule using fields such as
CommandLine|containsorParentImage, then verify the converted Splunk, KQL, and Elastic queries use the actual mapped field names.
Network and DNS investigations
- Generate an approved connection and confirm Event ID 3 records the process responsible, not merely the host connection.
- Validate whether IP addresses, ports, protocol, and the
Initiatedvalue survive normalization. NAT, proxies, and IPv6 can affect interpretation. - Use Event ID 22 to compare the querying process with DNS logs from the resolver or network sensor. A DNS event alone does not prove a successful connection.
- In a SIEM, test a correlation that joins process creation, DNS, and network events by host and process context where available.
File, registry, driver, and process-access investigations
- For file creation, distinguish Event ID 11 from deletion events. Confirm whether hashes are available for the file type and collection path being tested.
- For registry detections, test create, value-set, and rename behavior separately. Event IDs 12, 13, and 14 have different semantics and should not be collapsed into one generic rule.
- Review Event ID 6 and signature fields for driver monitoring. Signature validation may be absent or inconclusive, so preserve the raw status and avoid treating missing data as trusted.
- For Event ID 10, validate source and target process fields, granted access, and exclusions. Process-access telemetry can be high volume; narrow rules using unusual source processes, target processes, or access combinations.
Example query mappings
A Sigma rule might select EventID: 1 and a suspicious parent-child relationship. In Splunk, the equivalent commonly begins with EventCode=1 or a normalized event.code=1, depending on the add-on. Sentinel KQL may use EventID == 1 in a custom Sysmon table or EventID == "1" in a string-mapped column. Elastic rules often use event.code: "1" with ECS fields such as process.command_line. These are mapping patterns, not universal queries: inspect one raw event and one normalized event before publishing a rule.
What to double-check
- Configuration scope: Confirm the rule includes the event type and that include, exclude, and condition logic behave as intended. A broad exclusion can silently remove evidence.
- Raw versus normalized data: Compare the original Windows event with the SIEM document. Record exact field names, data types, casing, arrays, and timestamp format.
- Time and identity: Check UTC handling, host naming, user format, process ID reuse, and whether collector delays affect joins.
- Hashes and signatures: Treat hash and signature fields as optional evidence. Document when they are missing and how detections behave in that case.
- Volume: Measure event rates for image-load, network, file, and process-access telemetry before enabling broad collection. Tune with exclusions based on stable, understood behavior.
- Alert behavior: Confirm the rule produces the expected alert, severity, investigation fields, and suppression behavior. A query that returns results is not necessarily an operational detection.
Use a validation worksheet with these columns: scenario, host, test timestamp, expected event ID, observed event ID, expected fields, observed fields, raw event location, normalized field mapping, query result, alert result, owner, and remediation date. This makes gaps visible during a SIEM migration or endpoint-agent upgrade.
Common mistakes
- Assuming an event ID is collected because the Sysmon service is running.
- Writing detections against display labels instead of stable, verified field names.
- Testing only on a workstation while deploying the configuration to servers with different software and traffic patterns.
- Using command-line fragments without parent process, signer, path, user, or network context.
- Ignoring exclusions that were added to control volume but now remove investigation evidence.
- Combining multiple event types into one rule without preserving their different meanings.
- Declaring coverage after an endpoint alert appears without checking whether the SIEM received the same evidence.
Keep tests safe and reversible. Use disposable hosts, approved utilities, internal destinations, and harmless files or registry values. Do not use real credential-dumping tools or destructive payloads to prove that telemetry works; a controlled diagnostic action is usually sufficient to validate collection and field mapping.
When to revisit
Re-run this checklist before seasonal planning cycles, after changing Sysmon configuration syntax or exclusions, when onboarding a new host class, during a SIEM migration, and after upgrading Windows, Sysmon, an endpoint agent, or a parser. Revisit it whenever an investigation reveals missing process, DNS, network, registry, or file evidence.
For a practical quarterly telemetry audit, select one benign test per scenario, capture raw and normalized events, execute each production detection, and compare expected versus observed results. Update the configuration, field-mapping notes, Sigma conversions, Splunk searches, KQL detections, and Elastic rules together. Track unresolved gaps as engineering work rather than silently weakening the detection. For related exercises, see the lateral movement emulation lab, the PowerShell detection lab, and the detection coverage matrix template.