Beyond characters: a machine learning approach to password strength analysis
Mrinmoy Borah, Afsana Laskar · 2025
With the ever-increasing reliance on digital systems, the importance of robust password security has become paramount. Passwords function as a barrier against unauthorised access, thwarting cybercriminals from commandeering our accounts and potentially causing financial harm or identity theft. This study investigates the strength of passwords using several classification machine learning techniques. We have implemented a hybrid methodology for feature engineering, integrating TF-IDF with a structural analysis of each password. Three novel algorithms—K Nearest Neighbors (KNN), Convolutional Neural Network (CNN), and Ensemble Neural Network (ENN)—alongside with conventional machine learning classification techniques, to assess password strength, have been introduced. Of the nine machine and deep learning models, seven demonstrated approximately 99% accuracy.