How robust is that identification? Does it just look for file hashes or identical character shapes? I imagine it is trivial to repackage a font file to break the hash fingerprint.