2026-10-30
Proving a sink safe from its callers — and the bug a mutation test caught
How cross-file call-site analysis cut C# SQL false positives on Juliet from 48.6% to 27.3%, after a first version that was faster to improve and wrong.
2026-10-30
When the benchmark label is wrong
XXE in Nokogiri, CSRF without a cookie, and rand() as attacker input: three cases where we score lower on purpose, and how we decide.
2026-10-30
Taint is not a boolean
Why our engines track a set of vulnerability classes per value, how sanitizers clear only what they neutralise, and what a numeric conversion keeps.