Compliance mapping

Buyers and receiver makers answer to public resilience frameworks. This folder maps what Kshana computes and records to those frameworks, row by row, and kshana compliance-report fills the mapping from the runs you actually have.

Wording. Kshana outputs support evidence for a requirement. They are not a rating or approval of any product, and none of these files is a statement from an assessment body; Kshana is not one. Requirement text is paraphrased, never reproduced.

Each file lists its source documents with version, URL, date read and what was and was not checked. Sources were read on 2026-10-09 (the EASA bulletin and the EN 16803 catalogue pages were checked again on 2026-10-10). Where a source could only be read through a mirror or a catalogue page, the file says so.

The report#

kshana compliance-report spoof-detect.result.json integrity-raim.result.json session.result.json

writes compliance-report.compliance.md and compliance-report.compliance.json (--out <base> changes the name). The inputs are result files from earlier runs: a scenario run writes <name>.result.json beside <name>.toml, and kshana receiver-trust writes its own. The kind of each run is read from the scenario file beside the result, from the content of a receiver-trust result, or from a top-level kind, and must be a kind the engine knows. A result is listed as not used, with the reason, and counts for nothing when its kind cannot be found, is not a known kind, or the sibling scenario and the result's own kind disagree. (The result's scenario_hash and a hash of the scenario file are not compared: the engine computes the first over its parsed scenario and a file hash is over the text, so they are different quantities.)

A kind label alone is not evidence. A run counts for a capability only when its result also carries the fields that kind writes (for example a jamming result needs its figures of merit and per-epoch rows; an integrity result needs its sample counts). The fields are listed in src/compliance/mapping.rs (KIND_FIELDS) and a result missing one is not counted. A hash that is not hexadecimal is not reported as a hash. A receiver-trust result whose counts are not whole numbers, or whose detected or agreeing count exceeds its evaluable count, is excluded and listed as not used.

For each row the report gives one of:

The gap column is printed for every row, including evidenced ones: a run that supports evidence for a capability still leaves most of a requirement to other evidence. The report repeats the wording rule at its head. --mapping prints the static tables and --sources the source list.

Receiver-trust runs also contribute the counts they state (events detected of those evaluable, predictions agreeing of those evaluable). Those describe that log, scored against events and tolerances stated before the run, and are reported as the run states them.

Capabilities#