Research On Surveillance Video Authentication Based On Perceptual Hash Algorithm

With the rapid development of computer and multimedia application technology,network monitoring cameras are widely applied as a valid measure,which provide people with social safety.It is a challenge for researchers to test whether a surveillance video is tampered by the illegal personnel during the process of transmission or storage,so as to verify the reliability and authenticity of the video.Perceptual hash algorithm has become one of the mainstream technology of authentication.It has been deeply researched and applied to different kinds of multimedia.Different from watermarking algorithm which embedded the secret key into an authenticated data,perceptual hash algorithm extracts a secret key from an authenticated data and stores it as additional information without affecting the original content of the data.Especially in the field of image authentication,the research of image perceptual hashing algorithm has attracted the attention of scholars and experts at home and abroad,and they have made many achievements.As the video can be regarded as a set of images with time information,we make use of the previous theoretical studies to improve the traditional video hash algorithm,such that we can verify real-time video streaming and fine-grained video whether has been tampered or not.This paper mainly focuses on the research of video transmission and storage authentication algorithm.Two main research work is as follows:(1)By using the user data-gram protocol,the application of live video can ensure the real-time of transmission.However,the unreliability of the protocol will make the transmitted video stream present the risk of noise interference and data loss.This will lead the secret key authentication algorithm based on video image frames to be trapped into the indistinguishable problem.Therefore,a new adaptive hashing algorithm based on video compression frames is proposed in this paper.According to data length of a frame,the algorithm divided it into a number of intervals with different step-size and extracted its spatial structure feature.The feature is quantized to generate the secret key.The experimental results show the proposed algorithm can meet real-time requirement,and owns good robustness and differentiation.Byanalyzing the similarity of the statistical video frames,it can effectively detect the video tampering.(2)The extremely challenging issue of multimedia security problem is the tamper detection of video picture frames(VPF),especially for fine-grained size tampering.To solve this problem,this pager proposed a video perceptual hashing algorithm which extracted feature on local patches with discrete wavelet transform(DWT)and singular value decomposition(SVD).The algorithm firstly conversed the VPF into scale and color space,and divided its three color channel into overlapping patches.With DWT transforming and SVD decomposing on each patch,the middle feature matrix can be generated.Then,the matrix can be compressed to abstract hashing,i.e.the hashing secrete key.It's feasible and effective to evaluate the time-spatial similarity of VPF by method of segmentation.Experimental results have demonstrated that the proposed algorithm owns better discriminability,and can detect fine-grained size tampering effectively.