Privacy Preserving And Application Of Image Convolution Algorithm

With the development of large data technology,we are in an era of information explosion.Image,video and other visual media information occupies an important position in our daily life.These large number of visual media information processing is also increasingly dependent on cloud computing,which led to leakage problems of users' privacy data that image,the video contained.In recent years,Exposure of a variety of privacy image leak events to people sounded the alarm,the image of the privacy protection more and more important.Privacy protection of computer vision algorithm has become a hot research field.The computer vision algorithm is developing rapidly,but most of the algorithms can't be separated from the image convolution calculation,so the use of image convolution calculation and security calculation to design a safe and efficient secret algorithm,with significant research value.1.Study the principle and application of image convolution.Based on the principle and application of image convolution,combined with the image convolution calculation,two kinds of secret convolution calculations based on random sub-image and random matrix are designed by several applications of image convolution.This work protected the user data privacy effectively,and completed of the convolution calculation correctly,preparing for privacy protection research on more complex computer vision algorithm.2.Depth study of face detection algorithm and secret face detection algorithm.Focusing on Viola & Jones face detection algorithms and secret face detection algorithms based on oblivious transfer protocol.The face detection algorithm based on the oblivious transmission protocol,combined with the Viola & Jones face detection algorithm and oblivious transmission protocol,proposed a secure point product protocol and a millionaire protocol,which is a safe and effective face detection algorithm.In order to solve the problem of low efficiency of secret face detection algorithm based on oblivious transfer protocol,a secret face detection algorithm based on random sub-image is designed.The client uses the algorithm to divide the original image into 256 binary sub-images,each of which has independent weights,and the sub-images are randomly sent to the server for detecting.Both the client and the server introduce a random number mechanism that achieves privacy that protects both the privacy of the client image data and the server parameters,and improves the detection rate by about 500 times more than face detection algorithm based on oblivious transmission protocol.3.Aiming at the problems of face detection algorithm based on random sub-image,this paper presents a secret face detection algorithm based on random matrix,which is inspired by image information hiding technique.Feature image pixels hidden in the irreversible matrix,but also the server parameters hidden in another irreversible matrix.Protect their data privacy by introducing a random reversible matrix.The final feature weight is obtained by matrix multiplication,and the feature weight is hidden in the matrix.In order to further reduce the computational complexity of random matrices,a secret face detection algorithm based on random vector is designed.The algorithm introduces the stochastic reversible matrix,the n-dimensional {0,1} vector and random number to ensure the security of the algorithm.These three secret face detection algorithms are based on image convolution of the privacy protection algorithm to protect the privacy of both data.Compared with the secret face detection algorithm based on oblivious transmission protocol,greatly improve the detection rate and maintain the original correct rate.