부산대학교 도서관

Passwords (and login credentials) are pivotal in our daily online activities, securing ‘things’ including emails and e-wallets. Yet, users tend to recycle similar passwords across multiple services, posing a significant security threat. The prevalence of easily guessed passwords exacerbates the security issues. This study proposes a machine learning model leveraging Term Frequency – Inverse Document Frequency (TF-IDF) scores that helps develop stronger passwords by analyzing how frequently specific characters appear in passwords. The proposed methodology involves utilizing logistic regression and the 000WebHost leaked password wordlist for training and validation. The outcomes of the machine learning model accurately assess users’ password strength, potentially bolstering the security of their diverse online accounts.