# Sigma: requests in web access logs

Use `ai_credential_file_request.yml` for the selected credential filename and `ai_configuration_file_request.yml` for the lower-confidence configuration group. Both use the standard `category: webserver` and the conventional `cs-uri-stem` field. The author is in `author`; no telemetry product is invented.

1. Collect access logs from the server or reverse proxy that receives the relevant traffic. Retain failed requests as well as successful responses.
2. Select your actual web-log dataset in your Sigma conversion pipeline. Map `cs-uri-stem` to the request path only. IIS `csUriStem` is a direct mapping; in a source with a combined target, split the literal `?` first. `cs-uri-query` alone is not the path.
3. Convert with your SIEM's supported backend and inspect the generated field names and case behavior. Default Sigma string matching is case-insensitive. An `endswith` value must keep its leading slash and exact filename ending.
4. In a test environment you control, make a request for an absent selected file, such as `/.claude/.credentials.json`, and confirm the ingested path matches. It is safe for that test to return `404`; no real credential or public decoy file is needed. Check an ordinary path and a query-only filename mention do not match.
5. Tune approved scans and intentionally published examples by server, source, or approved test window. Keep raw request evidence available to an authorized responder.

Only the path is required to match; missing status or client fields should not suppress the request. Missing path means no match. Optional response fields in `fields` are investigation context, not extra selection conditions.

No URL decoding is done in the YAML. A raw `%2eclaude` path will not match the literal `.claude` rule. Decide and document your ingestion mapping; use the native queries when one additional decode pass fits your logs. The rules do not cover every alternative filename or server normalization behavior.

The exact YAML parses with pySigma and converts to the Splunk backend in a local field-mapping check. This checks syntax and generated semantics, not a real log source or tenant. Platform searches are provided separately with explicit source selection and normalization.

Primary references: [Sigma log sources](https://sigmahq.io/docs/basics/log-sources.html), [webserver field taxonomy](https://sigmahq.io/sigma-specification/specification/sigma-appendix-taxonomy.html), and [string modifiers](https://sigmahq.io/sigma-specification/specification/sigma-appendix-modifiers.html).
