An offline attack is a guessing attempt against stolen hashes, encrypted files, or other copied data without talking to the live service. Because the attacker can work on a local copy, there are no login locks, rate limits, or alerts from the target system. This makes offline attacks especially dangerous after a breach that exposes password hashes, token material, or poorly protected secrets.
In practice, attackers use dictionaries, password rules, and GPUs to test huge numbers of guesses quickly. Weak hashes, unsalted hashes, and fast algorithms such as older password schemes are easier to crack; slow, salted functions such as bcrypt, scrypt, or Argon2 raise the cost. Defenders reduce offline risk by hashing secrets with strong algorithms, unique salts, and high work factors, and by assuming any leaked hash set may eventually be tested offline.