On February 13, 2026, Lukás Pleva successfully defended his MSc thesis entitled „Evaluation of multi-modal biometric authentication in multi-party computation“.
Abstract:
Biometric authentication is convenient because a user does not need to remember complex passwords nor does he/she need to depend on a specific device to keep their cryptographic keys. To authenticate, one needs to compare features from a previous biometric sample with a newly sampled one. These biometric samples are considered sensitive personal data, and their leakage can have irreversible consequences for the individual.
Therefore, the protection of a stored biometric reference is very important.
For this reason, it is desirable for no single entity to know what a data subject’s biometric reference looks like. Additionally, only comparison score or binary result should be revealed to the system during a biometric verification. When designing any system for the future, it is important to consider the upcoming threats. A well known threat to the current cryptography is the development of quantum computers. Such a computer can create new possibilities, but this also includes the option to break the security of commonly used public-private key encryption schemes. For this reason, an evaluation of quantumsafe methods for distributed storage and distributed distance computation in biometric verification has been conducted in this work. A common way to increase the security of biometric authentication is to use multiple modalities. Therefore, this work investigates common approaches for a modality fusion and their applicability to such authentication system. This research resulted in the design of a multiparty protocol and its implementation, in
cluding an evaluation of its biometric and computational performance. The proposed sys
tem achieves the same biometric performance as the original plaintext nonquantized scorefused multimodal distance computation. Moreover, its computational performance outperforms similar existing solutions in many common scenarios. To the best of the author’s knowledge, this is the only system that scales to an arbitrary number of parties while preserving the biometric performance.