Reflexive Memory Authenticator: A Proposal for Effortless Renewable Biometrics

Abstract

Today's biometric authentication systems are still struggling with replay attacks and irrevocable stolen credentials. This paper introduces a biometric protocol that addresses such vulnerabilities. The approach prevents identity theft by being based on memory creation biometrics. It takes inspiration from two different authentication methods, eye biometrics and challenge systems, as well as a novel biometric feature: the pupil memory effect. The approach can be adjusted for arbitrary levels of security, and credentials can be revoked at any point with no loss to the user. The paper includes an analysis of its security and performance, and shows how it could be deployed and improved.

Presentation attack detection and biometric recognition in a challenge-response formalism

“…Challenges can also vary in degree of conscious control someone has over starting and/or halting processing the challenge in the scope of the biometric system. Users can have full control over the response, e.g., speaking a challenge word [14], or the response can be mostly automated and/or reflexive, e.g., pupil dilation changes in response to certain stimuli [15].…”

“…Other time, the security of the challenge-response protocol relies strongly on the user having had some form of practice during initial phases of application use. In [15], patterns in pupil dilation are used for verification, which arise from the presentation of a sequence of novel and non-novel pictures. It is possible to determine (non-)novelty, because users were presented with a sufficiently individualized selection of pictures during an enrollment phase.…”

“…This in turn should result in articles that describe protocols that are very different in kind. What follows is a discussion of 4 articles [6,12,15,23]. They show the diversity and creativity in types of challenge-response protocols that are currently being researched (and used), and how this diversity is represented in the space spun by our categorization.…”

“…In this method, users are shown a series of pictures in the enrollment phase [15]. During verification, users are shown a selection of familiar pictures from the enrollment phase, as well as unfamiliar pictures in rapid succession (the challenge).…”

“…An interesting experiment is to take existing challenge-response protocols and change their categorization in one of the five dimensions to use as a starting point for new challenge-response protocols, e.g., we can start with the image familiarity [15] method and imagine it using a population-wide model, rather than a single-subject model. Enrollment would then consist of a set of images that is the same for each user, which they are trained on.…”