基于随机子图表示的云环境人脸图像隐秘检测  

Private Face Detection Based on Random Sub-images in Cloud

作　　者：袁鹏[1,2] 金鑫[1] 李晓东[1] 赵耿[1] 吴亚明[1,2] 马铭鑫 田玉露[1] 陈迎亚[1] 

机构地区：[1]北京电子科技学院,北京100070 [2]西安电子科技大学,西安710071

出　　处：《系统仿真学报》2016年第9期2195-2200,共6页Journal of System Simulation

基　　金：国家自然科学基金(61170037;61402021);中央高校基本科研业务(2015XSYJ25);国家档案局科技计划(2015-B-10);虚拟现实技术与系统国家重点实验室开放课题(BUAA-VR-16KF-09)

摘　　要：为了在云端服务器中进行终端用户的人脸图像检测的同时能够保护双方的数据隐私,提出了一种基于随机子图表示的人脸图像隐秘检测方法。终端将原图像根据随机子图产生算法划分为2值的子图像加权和,将子图像的权值随机排序,将子图像按照权值随机排列的顺序发给云端服务器。云端服务器利用其人脸检测算法对子图像进行检测,终端根据随机子图分解合并检测结果。云端和终端都引入了随机数机制,以保护终端的人脸图像信息和云端的人脸检测算法参数。该方法的创新点是提出了随机子图算法并应用于图像的人脸检测中,使云端隐私保护的人脸检测得以实现。In order to detect faces of terminal face image in the cloud at the same time protect both privacy of data,a method of face images privacy detection based on random sub-Images representation was proposed. Terminal divided original image into 2 value sub-images weighted sum based on random sub-images generation algorithm and randomly arranges weights of sub-images. Terminal sent sub-images according to the weights of random sequence to the cloud server. Cloud server detected sub-images with its face detection algorithm. Terminal merges test results based on random sub were exploded. Two random vectors were leveraged to protect the parameters of the trained face detector and the content of the terminal image. The innovation of this method is that the random sub-images thought is proposed and is applied to the image face detection making privacy face detection achieve in the cloud.

关 键 词：云端 人脸检测 隐秘 随机子图 随机数 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]