Biometric authentication systems increasingly rely on privacy-preserving mechanisms to protect sensitive templates, such as encrypted or transformed representations of fingerprints, faces, or irises. But a new preprint on arXiv argues that the authentication interface itself may undermine these protections.
The paper, titled "How Much Feedback Is Needed to Break Template Protection? The Geometric Vulnerability of Biometric Authentication," identifies a geometric vulnerability. Even when templates are stored in a protected form, the responses returned during authentication attempts can reveal information about the underlying biometric data. The authors analyze how an attacker could exploit this feedback to reconstruct the original template.
Because the source is a single preprint, the findings have not yet been peer-reviewed. The abstract does not specify the exact amount of feedback required, but it frames the problem as a quantitative one: the less feedback needed, the more severe the risk. The work highlights a gap between protecting stored templates and protecting the live authentication process itself.