Lip Reading-Based User Authentication Through Acoustic Sensing on Smartphones
Lip Reading-Based User Authentication Through Acoustic Sensing on Smartphones
复制标题
通过智能手机上的声学传感进行基于唇读的用户身份验证
DOI：
10.1109/tnet.2019.2891733
复制
发表时间：
2019-02-01
影响因子：
3.7
通讯作者：
Li, Minglu
中科院分区：
文献类型：
--
作者：
Lu, Li;Yu, Jiadi;Li, Minglu
为了防止用户隐私泄露，越来越多的移动设备采用基于生物特征的认证方法，如指纹、人脸识别、声纹认证等，以加强对隐私的保护。然而，这些方法容易受到重播攻击。尽管最先进的解决方案使用活体验证来对抗攻击，但现有的方法对周围环境很敏感，如周围的灯光和周围的可听噪音。为此，我们探索了利用用户的嘴巴运动来验证用户身份验证的活跃性，这对噪声环境是鲁棒的。本文提出了一种基于唇读的用户认证系统LipPass，该系统通过智能手机上的声音感知来提取用户说话嘴巴独特的行为特征，用于用户认证。我们首先研究了由用户说话的嘴巴引起的声音信号的多普勒分布，发现不同的人有不同的嘴巴运动模式。为了刻画嘴部运动的特征，我们提出了一种基于深度学习的方法来从多普勒图像中提取有效的特征，并利用Softmax函数、支持向量机和支持向量域描述分别构造用于嘴部状态识别、用户识别和欺骗检测的多类识别器、二进制分类器和欺骗检测器。然后，我们开发了一种平衡的基于二叉树的认证方法，利用这些二进制分类器和欺骗检测器针对注册用户准确地识别每个人。通过在4个真实环境中对48名志愿者的广泛实验，LipPass在用户识别和欺骗检测方面的准确率分别达到了90.2%和93.1%。
To prevent users' privacy from leakage, more and more mobile devices employ biometric-based authentication approaches, such as fingerprint, face recognition, voiceprint authentications, and so on, to enhance the privacy protection. However, these approaches are vulnerable to replay attacks. Although the state-of-art solutions utilize liveness verification to combat the attacks, existing approaches are sensitive to ambient environments, such as ambient lights and surrounding audible noises. Toward this end, we explore liveness verification of user authentication leveraging users' mouth movements, which are robust to noisy environments. In this paper, we propose a lip reading-based user authentication system, LipPass, which extracts unique behavioral characteristics of users' speaking mouths through acoustic sensing on smartphones for user authentication. We first investigate Doppler profiles of acoustic signals caused by users' speaking mouths and find that there are unique mouth movement patterns for different individuals. To characterize the mouth movements, we propose a deep learning-based method to extract efficient features from Doppler profiles and employ softmax function, support vector machine, and support vector domain description to construct multi-class identifier, binary classifiers, and spoofer detectors for mouth state identification, user identification, and spoofer detection, respectively. Afterward, we develop a balanced binary tree-based authentication approach to accurately identify each individual leveraging these binary classifiers and spoofer detectors with respect to registered users. Through extensive experiments involving 48 volunteers in four real environments, LipPass can achieve 90.2% accuracy in user identification and 93.1% accuracy in spoofer detection.