An Intelligent Machine Learning Framework for Password Strength and Security System

Fuente: Zenodo
Salvato in:
Dettagli Bibliografici
Autori principali: Davies Isobo Nelson, Cookey Ibiere Boma, Godspower Oraye, Deedam Futune B.
Natura: Recurso digital
Lingua:Antico inglese
Pubblicazione: Zenodo 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866902289152737280
author Davies Isobo Nelson
Cookey Ibiere Boma
Godspower Oraye
Deedam Futune B.
author_facet Davies Isobo Nelson
Cookey Ibiere Boma
Godspower Oraye
Deedam Futune B.
contents <p>Password-based authentication remains the primary security mechanism for protecting user accounts and sensitive information, yet weak password selection continues as a pervasive global vulnerability. Traditional rule-based password strength meters exhibit significant limitations in adapting to emerging attack strategies and identifying subtle vulnerability patterns. This study presents an intelligent Long Short-Term Memory (LSTM) based framework for password strength classification that addresses these shortcomings through advanced Deep Learning techniques. The proposed system integrates a bidirectional LSTM network with comprehensive feature extraction pipelines incorporating Shannon entropy calculations, pattern recognition metrics, and statistical analysis to assess password security across weak, medium, and strong classifications. The framework was trained and validated on a diverse dataset of 1,000,000 passwords compiled from breach databases and synthetic sources, partitioned into 70% training, 15% validation, and 15% testing sets. The experimental results for the system demonstrate exceptional performance with 96.8% accuracy, outperforming traditional rule-based systems by 23.4 percentage points and commercial password meters by 12.1-18.5 percentage points. The system achieves 98.3% recall for weak password detection with only 1.7% false negative rate, compared to 8.8-13.9% for commercial solutions. Field deployment with 500 users over 30 days validated practical impact, showing 40.7% higher password entropy, 204.3% improvement in strong password adoption, and 69.5% reduction in dictionary word usage. The lightweight architecture (4.7 MB model size, 1.8ms inference time) enables real-time integration into web applications and enterprise authentication systems. The framework provides actionable, context-aware security recommendations and achieves detection rates exceeding 94% across diverse attack vectors including dictionary attacks, keyboard patterns, sequential characters, and previously breached credentials. The study demonstrates that LSTM-based approaches significantly enhance password security assessment capabilities and effectively guide users toward adopting stronger authentication practices.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_19915091
institution Zenodo
language ang
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle An Intelligent Machine Learning Framework for Password Strength and Security System
Davies Isobo Nelson
Cookey Ibiere Boma
Godspower Oraye
Deedam Futune B.
Password Security, LSTM, Deep Learning, Authentication, Cybersecurity, Machine Learning, Password Strength Classification
<p>Password-based authentication remains the primary security mechanism for protecting user accounts and sensitive information, yet weak password selection continues as a pervasive global vulnerability. Traditional rule-based password strength meters exhibit significant limitations in adapting to emerging attack strategies and identifying subtle vulnerability patterns. This study presents an intelligent Long Short-Term Memory (LSTM) based framework for password strength classification that addresses these shortcomings through advanced Deep Learning techniques. The proposed system integrates a bidirectional LSTM network with comprehensive feature extraction pipelines incorporating Shannon entropy calculations, pattern recognition metrics, and statistical analysis to assess password security across weak, medium, and strong classifications. The framework was trained and validated on a diverse dataset of 1,000,000 passwords compiled from breach databases and synthetic sources, partitioned into 70% training, 15% validation, and 15% testing sets. The experimental results for the system demonstrate exceptional performance with 96.8% accuracy, outperforming traditional rule-based systems by 23.4 percentage points and commercial password meters by 12.1-18.5 percentage points. The system achieves 98.3% recall for weak password detection with only 1.7% false negative rate, compared to 8.8-13.9% for commercial solutions. Field deployment with 500 users over 30 days validated practical impact, showing 40.7% higher password entropy, 204.3% improvement in strong password adoption, and 69.5% reduction in dictionary word usage. The lightweight architecture (4.7 MB model size, 1.8ms inference time) enables real-time integration into web applications and enterprise authentication systems. The framework provides actionable, context-aware security recommendations and achieves detection rates exceeding 94% across diverse attack vectors including dictionary attacks, keyboard patterns, sequential characters, and previously breached credentials. The study demonstrates that LSTM-based approaches significantly enhance password security assessment capabilities and effectively guide users toward adopting stronger authentication practices.</p>
title An Intelligent Machine Learning Framework for Password Strength and Security System
topic Password Security, LSTM, Deep Learning, Authentication, Cybersecurity, Machine Learning, Password Strength Classification
url https://doi.org/10.5281/zenodo.19915091