Beyond Credential Stuffing - Password Similarity Models Using Neural Networks

Explore cutting-edge research on password security in this IEEE Symposium presentation. Delve into the development of advanced password similarity models using neural networks, trained on a vast dataset of 1.4 billion leaked credentials. Discover how these models enable a highly effective targeted attack that compromises over 16% of user accounts in under 1,000 guesses, even with state-of-the-art countermeasures in place. Learn about the practical implications through a case study involving a large university authentication service. Examine the innovative defense strategy of personalized password strength meters (PPSMs), designed to warn users about vulnerable password choices based on their previously compromised credentials. Gain insights into the compact and deployable PPSM solution that accurately estimates password strength against known guessing attacks, all compressed into less than 3 MB.

Beyond Credential Stuffing: Password Similarity Models using Neural Networks