Litcius/Paper detail

Password Strength Analysis and its Classification by Applying Machine Learning Based Techniques

Sakya Sarkar, Mauparna Nandan

20222022 Second International Conference on Computer Science, Engineering and Applications (ICCSEA)15 citationsDOI

Abstract

Passwords serve as a vital authentication mechanism for any system’s protection. Although numerous more secure methods of authentication exists such as biometrics and smart cards but still the largely popular technique is password authentication for ensuring protection of any system on account of its ease of implementation. But, passwords are exposed to various types of vulnerabilities due to the predictable patterns which human employs for setting passwords likely dictionary words, commonly used people and places names, keyboards patterns, date of birth, known phrases etc. and thereby, pave a route for online or offline muggers to intrude into the system by assuming the passwords. For guessing the passwords, various password cracking tools are available either online or offline and most of the tools can readily crack such accounts possessing passwords with weak strengths or passwords with common patterns. Therefore, the organizations should enforce unassailable strategies governing the use of strong passwords such that those vulnerabilities can be addressed. Hence, the most efficient way to protect the system against these online and offline attacks is by enforcing the users to implement a strong password. In this proposed work, prediction of the strength of a password has been modeled as a classification task with the aid of multiple supervised machine learning algorithms which were used for the learning purpose. The novelty of this work lies in the fact, that two new algorithms namely, XGBoost and Multilayer Perceptron have been implemented to determine the strength and the corresponding category of a password which has not been evaluated in similar other works. However, during testing the models with the test dataset, it has been observed that XGBoost outperformed the other machine learning classifiers with an accuracy of 99%. The final results reveal that machine learning approaches possesses ample capability to classify the passwords into different categories, namely, Weak, Medium and Strong passwords.

Topics & Concepts

PasswordComputer scienceArtificial intelligenceMachine learningPattern recognition (psychology)Computer securityUser Authentication and Security SystemsAdvanced Malware Detection TechniquesAdvanced Steganography and Watermarking Techniques