Development of a Password Strength Evaluation System for Enhanced Security Using Deep Learning Based Approach

ABSTRACT
Traditional password strength evaluation systems, which rely on rule-based heuristics, often fail to detect sophisticated patterns that can be exploited by attackers. This research addresses these limitations by developing a deep learning-based password strength evaluation system capable of providing more accurate assessments and actionable feedback to users. The study employs a hybrid neural network architecture incorporating LSTM (Long Short-Term Memory) layers to capture sequential patterns in passwords, enhancing the system's ability to identify vulnerabilities that conventional methods miss. The research methodology involved collecting and preprocessing password datasets, developing and training the deep learning model, and implementing a user-friendly interface for real-time password evaluation. The system was evaluated using comprehensive metrics including accuracy, precision, recall, and F1-score, with results demonstrating significant improvement over traditional rule-based approaches. Character-level tokenization enabled the model to detect subtle patterns that contribute to password vulnerability, while the feedback mechanism provided users with specific recommendations to enhance their password security. Key findings revealed that the deep learning model effectively learned complex password patterns without explicit feature engineering, with experimental results showing balanced performance across different password strength categories. The system successfully bridges the gap between technical security metrics and practical user guidance, promoting better password creation habits. This research contributes to the field of cybersecurity by demonstrating how deep learning techniques can enhance password security evaluation, offering a flexible and adaptive approach that evolves with emerging threats.