Lyric-based passwords: Enhancing security and recall with AI

Abstract

In the digital age, text-based passwords remain the cornerstone of user authentication. However, the balance between security and memorability remains a significant challenge. Users often face a dilemma between creating complex passwords that are difficult to remember and simpler ones that are vulnerable to attacks.

This research introduces a novel approach to password generation by leveraging linguistic patterns from song lyrics and advanced machine learning models. By processing over 5 million lyrics from the AZ Lyrics and Genius datasets, we identify memorable linguistic constructs, such as verb phrases, to create secure and user-friendly passwords. Transformer architectures are employed for password generation, while LSTM-based models assess their security.

A web application integrates these features to enhance usability, offering mnemonic aids such as narrative generation and interactive tools for real-time password creation. This system educates users on best practices and simplifies password management through an engaging interface. Comparative studies demonstrate that lyric-based passwords outperform traditional recall and security metrics methods. By balancing usability and robustness, this approach sets a new standard for password management systems and offers a forward-thinking solution to a persistent cybersecurity challenge.