Abstract:
Passwords provide the first line of defense against unauthorized access to your computer
and personal information. Though there are many alternatives to passwords for access
control, a password is the more compellingly authenticating the identity in many
applications. They provide a simple, direct means of protecting a system and they
represent the identity of an individual for a system. So, life these days have become
largely dependent on password for many purposes. Logging in to computer accounts,
retrieving email from the server, transferring funds, online shopping, accessing programs,
databases, networks, websites, and even reading the morning newspaper online. The
problem of selecting and using good passwords is becoming more important every day.
In this work password strength prediction is modeled as a classification task and
supervised machine learning techniques were employed. Widely used supervised
machine learning techniques namely Logistic Regression, Naïve Bayes Classifier,
Support Vector Machine, and Gradient Boosting Algorithms were used for learning the
model. The proposed model was applied to two different datasets and states that this
model is stable in terms of accuracy. The results of the models were also compared with
the existing password strength checking tools. The findings show that the machine
learning approach has substantial capability to classify the extreme cases – Strong,
Medium, and Weak passwords.
©
