Date: October 30, 2026
Speaker: Alaa Daffalla, Ph.D. Student, Cornell Bowers
Title: A Framework for Abusability Analysis: The Case of Passkeys in Interpersonal Threat Models

 A color photo of a woman in cap and gown.

Abstract: The recent rollout of passkeys by hundreds of web services online is the largest attempt yet to achieve the goal of passwordless authentication. However, new authentication mechanisms can often overlook the unique threats faced by at-risk users, such as survivors of intimate partner violence, human trafficking, and elder abuse. Such users face interpersonal threats: adversaries who routinely have physical access to devices and either know or can compel disclosure of passwords or PINs. The extent to which passkeys enable or mitigate such interpersonal threats has not yet been explored. We perform the first analysis of passkeys in interpersonal threat models. To do so, we introduce an abusability analysis framework to help practitioners and researchers identify ways in which new features can be exploited in interpersonal threat models. We then apply our framework to the setting of passkeys, ultimately investigating 19 passkey-supporting services. We identify a variety of abuse vectors that allow adversaries to use passkeys to cause harm in interpersonal settings. In the most egregious cases, flawed implementations of major passkey-supporting services allow ongoing illicit adversarial access with no way for a victim to restore security of their account. We also discover abuse vectors that prevent users from accessing their accounts or that help attackers emotionally manipulate (gaslight) users.

Paper link: https://www.usenix.org/system/files/usenixsecurity25-daffalla.pdf 

Bio: Alaa Daffalla is a fourth-year Ph.D. in Computer Science student at Cornell University. She is based in NYC at the Cornell Tech campus. She studies the safety risks and harms resulting from emerging technologies. Her work has investigated the abusability of account security management tools and passwordless authentication. More recently, she has been exploring emerging threats from generative AI, including its role in facilitating the creation of SNCII and deepfakes. She also explores designing new interventions for abuse prevention and mitigation. To achieve this, she takee a mixed-methods approach that combines measurement studies, user research, red teaming, and safety auditing.