SmileAuth: Using Dental Edge Biometrics for User Authentication on Smartphones
SmileAuth: Using Dental Edge Biometrics for User Authentication on Smartphones
复制标题
SmileAuth：使用 Dental Edge 生物识别技术在智能手机上进行用户身份验证
DOI：
10.1145/3411806
复制
发表时间：
2020-09-01
期刊：
影响因子：
--
通讯作者：
Cao, Zhichao
中科院分区：
文献类型：
--
作者：
Jiang, Hongbo;Cao, Hangcheng;Cao, Zhichao
用户身份验证对于智能手机的安全和隐私保护至关重要。虽然在智能手机上有各种各样的认证方案，但security.aws一直在被发现。指纹传感器和反监视假面具可以欺骗面部识别。在本文中，我们提出了一种新的用户身份验证系统SmileAuth，利用人的牙齿边缘生物特征的可靠和方便的用户身份验证的独特功能。SmileAuth通过轻微移动智能手机从不同的摄像头角度捕获一些图像来提取一系列牙齿边缘特征。这些独特的特征是由齿的尺寸、形状、位置和表面磨损决定的。SmileAuth对图像spoo.ng、基于视频的攻击、物理强制攻击和假牙攻击具有鲁棒性。我们在Android智能手机上实现了SmileAuth的原型，并通过招募300多名志愿者对其性能进行了全面评估。实验结果表明，SmileAuth在不同场景下的总体准确率为99.74%，F-score为98.69%，FNR为2.31%，FPR为0.25%。另外两对双胞胎的实验表明，牙齿边缘生物特征是唯一的，足以有效地区分双胞胎。
User authentication is crucial for security and privacy protection on smartphones. While a variety of authentication schemes are available on smartphones, security.aws have been continuously discovered. Fingerprint.lms can deceive.ngerprint sensors and anti-surveillance prosthetic masks can spoof face recognition. In this paper, we propose a novel user authentication system SmileAuth that leverages the unique features of people's dental edge biometrics for reliable and convenient user authentication. SmileAuth extracts a series of dental edge features by slightly moving the smartphone to capture a few images from di.erent camera angles. These unique features are determined by the tooth size, shape, position and surface abrasion. SmileAuth is robust against image spoo.ng, video-based attack, physically forced attack and denture attack. We implemented the prototype of SmileAuth on Android smartphones and comprehensively evaluated its performance by recruiting more than 300 volunteers. Experimental results show that SmileAuth can achieve an overall 99.74% precision, 98.69% F-score, 2.31% FNR and 0.25% FPR in diverse scenarios. Additional experiments with two pairs of twins demonstrate that dental edge biometrics are unique enough to e.ectively distinguish twins.