Password Strength analysis and Improvement

Abstract

This thesis analyzes real-world password weaknesses using a selected subset of the RockYou dataset. It identifies common patterns, structural weaknesses, and the limits of entropy-only strength estimation. Based on these findings, it develops a password strength evaluation model that combines entropy, pattern detection, blacklist checking, crack-time estimation, and user feedback. The model was implemented as a working prototype and compared with existing tools such as zxcvbn and Chrome. The results show that a dataset-informed and pattern-aware approach can provide a more realistic and transparent assessment of password strength.