Kim Hanvit
Annu Int Conf IEEE Eng Med Biol Soc. 2017 Jul;2017:454-457. doi: 10.1109/EMBC.2017.8036860.
Biometrics such as ECG provides a convenient and powerful security tool to verify or identify an individual. However, one important drawback of biometrics is that it is irrevocable. In other words, biometrics cannot be re-used practically once it is compromised. Cancelable biometrics has been investigated to overcome this drawback. In this paper, we propose a cancelable ECG biometrics by deriving a generalized likelihood ratio test (GLRT) detector from a composite hypothesis testing in randomly projected domain. Since it is common to observe performance degradation for cancelable biometrics, we also propose a guided filtering (GF) with irreversible guide signal that is a non-invertibly transformed signal of ECG authentication template. We evaluated our proposed method using ECG-ID database with 89 subjects. Conventional Euclidean detector with original ECG template yielded 93.9% PD1 (detection probability at 1% FAR) while Euclidean detector with 10% compressed ECG (1/10 of the original data size) yielded 90.8% PD1. Our proposed GLRT detector with 10% compressed ECG yielded 91.4%, which is better than Euclidean with the same compressed ECG. GF with our proposed irreversible ECG template further improved the performance of our GLRT with 10% compressed ECG up to 94.3%, which is higher than Euclidean detector with original ECG. Lastly, we showed that our proposed cancelable ECG biometrics practically met cancelable biometrics criteria such as efficiency, re-usability, diversity and non-invertibility.
诸如心电图（ECG）之类的生物特征识别技术为验证或识别个人提供了一种便捷且强大的安全工具。然而，生物特征识别技术的一个重要缺点是它不可撤销。换句话说，生物特征一旦被泄露，实际上就无法再使用了。为了克服这一缺点，人们对可撤销生物特征识别技术进行了研究。在本文中，我们通过在随机投影域中的复合假设检验推导广义似然比检验（GLRT）检测器，提出了一种可撤销的心电图生物特征识别技术。由于可撤销生物特征识别技术通常会出现性能下降的情况，我们还提出了一种具有不可逆引导信号的引导滤波（GF），该引导信号是心电图认证模板的非可逆变换信号。我们使用包含89名受试者的ECG-ID数据库对我们提出的方法进行了评估。使用原始心电图模板的传统欧几里得检测器的检测概率为93.9%（误识率为1%时的检测概率），而使用10%压缩心电图（原始数据大小的1/10）的欧几里得检测器的检测概率为90.8%。我们提出的使用10%压缩心电图的GLRT检测器的检测概率为91.4%，优于使用相同压缩心电图的欧几里得检测器。使用我们提出的不可逆心电图模板的引导滤波进一步将使用10%压缩心电图的GLRT检测器的性能提高到了94.3%，高于使用原始心电图的欧几里得检测器。最后，我们表明我们提出的可撤销心电图生物特征识别技术实际上满足了可撤销生物特征识别技术的标准，如效率、可重用性、多样性和不可逆性。
