Privacy-Preserving Biometric Matching via Secure Two-Party Computation

This thesis presents a detailed exploration of Funshade, a framework designed to enable secure biometric authentication through privacy-preserving protocols. With biometric data increasingly used for security, protecting this sensitive information is cardinal. Funshade provides a method for comparing biometric data between parties without revealing the actual data itself, ensuring privacy and security.
As a particular example, the Digidow project uses biometric authentication to determine if the access can be given to the individual detected by a sensor. Digidow employs a decentralized structure in which each person's biometric template is stored either with a chosen cloud provider or on a personal server. Since biometric data is stored across potentially untrusted locations, the comparison of the stored template with the live data from a sensor requires a secure and privacy-preserving solution that protects the data even in the presence of potentially malicious parties. For this task Funshade is viewed as a potential candidate, as the participants are able to keep the sensitive data private that is needed for the verification.
A prototype is implemented in Rust, chosen for its strong memory-safety and performance features. Throughout the thesis, challenges such as managing Rust’s memory model, and optimizing cryptographic functions were addressed. Additionally, several areas for future improvement are identified. These enhancements aim to improve security, usability, and adaptability of the framework in diverse applications.