Fully Homomorphic Encryption Operators for Score and Decision Fusion in Biometric Identification
Fully Homomorphic Encryption Operators for Score and Decision Fusion in Biometric Identification
复制标题
DOI：
10.1109/wifs58808.2023.10374571
复制
发表时间：
2023-12
期刊：
影响因子：
--
通讯作者：
Tilak Sharma;Mahika Wason;Vishnu Naresh Boddeti;Arun Ross;N. Ratha
中科院分区：
文献类型：
--
作者：
Tilak Sharma;Mahika Wason;Vishnu Naresh Boddeti;Arun Ross;N. Ratha
生物特征融合的原理，它需要结合多个生物特征匹配器，通常用于（a）提高识别精度和（B）增加生物特征系统的安全性。然而，融合可以暴露由对手可以利用的个体生物特征匹配器生成的信息。本文探讨了利用全同态加密（FHE）进行评分级和决策级融合的可能性，以增强安全性和隐私性。在决策级和分数级融合的背景下，我们适当的比较算法，可以操作完全同态加密的输入。此外，对于分数级融合，我们在加密域中执行分数归一化，从而增强了分数数据的隐私性和安全性。在NIST BSSR 1数据集上的实验表明，FHE可以提供一种可行的解决方案，用于保护生物特征评分和决策数据，同时保留它们在融合中的实用性。本文的贡献如下：（a）利用和实现FHE兼容的生物识别框架中的操作;和（B）评估这样的框架在真实世界的数据集上的性能。
The principle of biometric fusion, which entails combining multiple biometric matchers, is often used to (a) improve recognition accuracy and (b) increase the security of biometric systems. However, fusion can expose information generated by individual biometric matchers that an adversary can exploit. This paper explores the possibility of performing score-level and decision-level fusion by utilizing fully homomorphic encryption (FHE) for enhanced security and privacy. In the context of decision-level and score-level fusion, we appropriate a comparison algorithm that can operate on fully homomorphically encrypted inputs. Furthermore, for score-level fusion, we perform score normalization in the encrypted domain, thereby enhancing the privacy and security of the score data. Experiments on the NIST BSSR1 dataset suggest that FHE can provide a viable solution for securing biometric scores and decision data while retaining their utility in fusion. The contributions of this paper are as follows: (a) leveraging and implementing FHE-compatible operations in a biometric identification framework; and (b) evaluating the performance of such a framework on a real-world dataset.