Publication Date: 2023/08/28
Abstract: Cybersecurity has become a serious threat to society because of the revolution on the internet. Due to the internet revolution worldwide people are consuming quintillion bytes of data on daily basis. The data consumption over the internet may increase in the feature at the same time the threats to internet security posing new questions to the world. One of the major problems in cybersecurity is image forgery. An effective mechanism to detect image forgery is needed to avoid complications in various fields like medical imaging, space research, defense, etc., where even small details in the images are very crucial. In the present research by taking the advantage of Artificial intelligence an effective model is built. This model in the pre-processing stage of the image uses superpixels. These features will be provided as inputs to the deep neural network. Basically, the neural network acts as a classifier of the images. The convolutional neural networks are built and optimized according to the input data. The convolutional neural networks are being trained by a large number of image data set and will be tested for the results. When the trained CNN is supplied with the images which are needed to be detected for the forgery in the initial stages the images will be divided into blocks that are non-uniform and features will be extracted which consists of superpixels. These features will be supplied to the classifier. The classifier not only detects the forged image and non forged image but also indicates the location of the forgery. The present research paper compares various methods of image forgery detection. In the comparison, the proposed method will enhance performance matrices in terms of accuracy, precession, Recall, etc.
Keywords: Forgery detection, Deep neural network, Artificial intelligence, Convolutional neural network, superpixels, Feature extraction, accuracy, precession, recall, confusion matrix.
DOI: https://doi.org/10.5281/zenodo.8289379
PDF: https://ijirst.demo4.arinfotech.co/assets/upload/files/IJISRT23AUG467.pdf
REFERENCES