Log redaction is a desired feature for PII data and as a complement to encryption at rest.
In its simplest form it could be implemented as a configuration option redactLogs orthogonal to the log level.
- Preserve enough info in the redaction function so that CEs et al. can do useful analytics on the output. This includes for query shapes for instance. This could be achieved through hashing actual data in the queries.
Things that could/should be redacted:
1. query parameters (things other than field names and dollarsign operators).
2. hostnames/IP addresses
3. User names maybe?
4. Namespace names maybe?
5. Error messages in case error messages include any of the above.
- duplicates
-
SERVER-12671 Provide option to allow "masking" of query variables within audit stream.
- Backlog
-
SERVER-17377 Allow removing attribute values from slow query logs, log query shape only
- Closed
- related to
-
SERVER-18946 I wish I can configure separately the profiler and log file
- Open