Identity by Design, Demographics by Accident: Demographic Leakage and Suppression in Behavioral Biometric Embeddings
Cryptography and Security
Summary
The authors studied how systems that use behavior patterns like voice or typing style to verify identity might accidentally reveal personal details like age or gender. They checked many different models and types of behavior data and tested ways to hide this extra information without stopping the system from recognizing users. They found that some behaviors, like voice, leak more private info but can be better protected, while others, like typing patterns, leak less but are harder to clean up. This shows there's a privacy risk in these systems that needs careful handling.
Authors
Iyadh Khan, Patrick Nilackshan, Mohamed Aathif, Mohamed Theesan, Sandareka Wickramanayake, Sanka Rasnayaka
Abstract
Behavioral biometric authentication (BBA) systems use deep learning models to transform biometric signals, such as eye movements, voice, keystroke/touchstroke dynamics, and gait, into identity embeddings for user authentication. While designed to encode identity, these embeddings may inadvertently reveal sensitive demographic attributes, including gender, age, and height. Consequently, an adversary with access to the authentication model can infer demographic information from biometric signals, including those of users unseen during training or enrollment. In this paper, we present the first systematic audit of demographic leakage in BBA systems, evaluating 11 models across 9 datasets spanning four biometric modalities. We further benchmark four post-hoc suppression methods-Incremental Variable Elimination (IVE), Hilbert-Schmidt Independence Criterion (HSIC), Adversarial Encoder-Decoder (AED), and Protected Attribute Suppression System (PASS) to assess their ability to mitigate demographic leakage while preserving authentication utility. Our analysis reveals substantial variation in leakage and suppressibility across modalities, model architectures, and learning objectives. While voice embeddings exhibit high leakage that can be effectively suppressed, keystroke/touchstroke embeddings exhibit lower leakage but are considerably more difficult to sanitize. These findings highlight a fundamental privacy risk in behavioral biometric authentication and provide insights into the factors governing demographic information suppressibility.