Bcrypt Password Hashing Silently Ignores Characters Beyond 72 Bytes, Breaking Verification Expectations
The bcrypt algorithm, widely used for password hashing in PHP and Python applications, processes only the first 72 bytes of any input password. Additional characters are ignored without error or warning during both hashing with password_hash and verification with password_verify. This behavior stems from the fixed 72-byte P-table size in the underlying Blowfish cipher as defined in the 1999 specification. Long passphrases, especially those using multibyte UTF-8 characters like Cyrillic or emojis, can result in completely different strings producing identical hashes. The limitation creates risks during password manager usage, system migrations, and scenarios involving shared prefixes. Modern libraries such as Python's bcrypt 4.x now explicitly reject passwords exceeding 72 bytes, while older implementations continue silent truncation. Recommended fixes include enforcing byte-length validation or pre-hashing with SHA-256 before bcrypt.
Habr•Vulnerabilities & Exploits
