railguard-benchmark

To benchmark security scan accuracy, store scan findings in a local SQLite database and compare them against ground-truth answer sheets to measure recall and precision across multiple repos and models.

Comparing security scan results across multiple runs is best done by persisting findings to a local SQLite database, enabling objective comparison of recall and precision between scan runs over time.

Yes, you need Python 3 and sqlite3 to run the scripts that store security scan findings in a local SQLite database and compare results against answer sheets or prior runs.

Yes, you can measure recall and precision for orchestrated, nofp, and other scan variants by storing their findings in SQLite and comparing them against ground-truth answer sheets.

Storing scan findings in SQLite improves benchmark reporting by providing auditability and enabling dashboards that track detector quality trends and compare accuracy across multiple runs.