Cancellable biometrics is the art of generating distorted or encrypted templates of original biometric templates. The evolution of cancellable biometrics is attributed to the advanced hacking technologies that can capture the original stored biometrics from databases. One of the solutions for this problem is to store cancellable biometric templates in the database rather than the original ones. This paper presents a cancellable face recognition scheme that is based on face image encryption with Fractional-Order (FO) Lorenz chaotic system. The basic idea is to generate user-specific random keys to be XORed with the red, green, and blue components of color face images. These keys are generated from the fractional-order Lorenz chaotic system. In addition, some post-processing is implemented on the encrypted color components of the face images with rotation and transposition of matrices. Finally, a wavelet fusion process is applied on these encrypted and processed face image components. The reason behind the utilization of wavelet fusion is to generate a single cancellable template for each color face image. Furthermore, wavelet fusion after post-processing of encrypted color components leads to a better degree of diffusion in the encrypted face templates. Actually, the encryption with the proposed algorithm is not full encryption, but it is appropriate for cancellable biometric applications. Moreover, the proposed scheme is secure due to the power of fractional-order Lorenz chaotic system that is very sensitive to initial conditions selected by the user. In addition, the post-processing incorporated with wavelet fusion is a non-invertible process. The validation of the proposed scheme is performed with experiments on FERET, LFW, and ORL databases. Evaluation metrics including Equal Error Rate (EER), Area under the Receiver Operating Characteristic (AROC) curve are utilized in the proposed scheme. Numerical values reveal EER levels close to zero and AROC values of 100% at low and mild noise levels.

可取消生物识别技术是生成原始生物识别模板的扭曲或加密模板的艺术。可取消生物识别技术的发展归功于先进的黑客技术，可以从数据库中捕获原始存储的生物识别技术。此问题的解决方案之一是将可取消的生物特征模板而不是原始模板存储在数据库中。本文提出了一种基于人脸图像加密和分数阶 (FO) Lorenz 混沌系统的可取消人脸识别方案。基本思想是生成用户特定的随机密钥，与彩色人脸图像的红色、绿色和蓝色分量进行异或。这些密钥是从分数阶洛伦兹混沌系统生成的。此外，通过矩阵的旋转和转置对人脸图像的加密颜色分量进行一些后处理。最后，对这些经过加密和处理的人脸图像组件应用小波融合过程。利用小波融合背后的原因是为每个彩色人脸图像生成单个可取消模板。此外，加密颜色分量后处理后的小波融合导致加密人脸模板中更好的扩散程度。实际上，所提出算法的加密不是完全加密，但适用于可取消的生物识别应用。此外，由于分数阶洛伦兹混沌系统对用户选择的初始条件非常敏感，因此所提出的方案是安全的。此外，结合小波融合的后处理是一个不可逆的过程。所提出方案的验证是通过在 FERET、LFW 和 ORL 数据库上进行的实验进行的。评估指标包括等错误率 (EER)、接收器操作特性 (AROC) 曲线下的面积在所提出的方案中使用。数值显示 EER 水平接近于零，AROC 值在低和轻度噪声水平下为 100%。