Image Authentication Method And System Based On Image Tampering Detection

With the popularization of computers and the development of network technology,there is deeper soil for rapid development and the large-scale spread of digital media content,which promotes cultural exchanges.However,due to its zero-cost copying and rapid dissemination,pirated products are widely population,which has seriously impacted the traditional field of copyright protection.In order to solve this problem,this paper proposes an algorithmic technical framework for image tampering detection and recognition.The algorithmic was used to detect additions,deletions,and remakes of images to protect original images.Based on the good effect of tampering detection on the original image,it has also achieved certain results in the detection of remake and deformed images.The main work of this paper is as follows:(1)A technical framework for tampered image detection is proposed.The framework combines Super Point with the BOW model to realize the near-duplicate image retrieval task,which is applied to find tampered images and original images.Combine the Super Point feature points of the original image and the tampered image with the Super Glue matching method,and apply it to the image affine perspective transformation to transform the tampered image and the original image to the same perspective for the next step of tampering detection.Compared with the traditional hand-designed and extracted features combined with BOW,the feature points extracted by Super Point are few and accurate,and the registration effect combined with Super Glue is good.Moreover,the extraction speed of Super Point is significantly improved compared with the artificially designed features,and the retrieval effect obtained by combining the BOW model is better.(2)A VGG-based residual network image forgery detection model is proposed.We added a Super Glue-based image registration module at the beginning of the model,in order to facilitate the development of subsequent tamper detection.Based on the existing VGG19 network model,we removed the fully connected layer and softmax layer of the last three layers of VGG19,retained the convolutional layer part of its feature extraction,and extracted the feature maps of different convolutional layers for upsampling.,splicing.The original image and the tampered image are paired into the network,and the L2 distance is calculated for the extracted feature map of the original image and the feature map of the tampered image,and the difference image of the two images is generated.It has a good effect on the tampering detection on the original image,and also has a good effect on the remake and deformed image.(3)Design and implement a blockchain-based image authentication system using the blockchain's Hyperledger Fabric framework and Flask framework.It provides a safe and efficient application system for image tampering detection and image authentication.