It’s a bespoke scanning setup designed to deal with GitHub’s scale, minimise false positives, and scan fast enough to be in the `git push` request/response cycle. Under the hood it’s using Intel’s hyperscan as the regex engine.

https://github.com/intel/hyperscan