Inter frame Tampering Detection based on DWT-DCT Markov Features and Fine tuned AlexNet Model

Nowadays, videos are simply captured and viral in little time, and video editing has turn out to be more convenient with editing software. Therefore, the validity of the videos becomes more important. Inter frame video counterfeiting is the most common type of video spoofing method that is difficult to detect with the naked eye. So far, it has been suggested that some algorithms detect Inter frame counterfeits based on artisanal characteristics, but the accuracy and processing speed of these algorithms remain a challenge. This article proposes Markov based approach to detecting this particular object. First, the unique Markov characteristics in the DCT domain are extended to capture not only the inter-block correlation but also the intra-block association among the block DCT coefficients. after that, supplementary features are built in the DWT domain to distinguish three types of dependencies between the wavelet coefficients across positions, scales, and orientations. After that, we will introduce a video tampering detection method to detect Inter frame video tampering based on Convolutional Neural Network (CNN) models by retraining the accessible CNN model trained on the ImageNet dataset. The proposed method is based on state-of-the-art CNN models, which are retrained to exploit the spatio-temporal relations in a video to strongly detect Inter frame fakes and we have also proposed a confidence score instead of the score of raw output based on these networks, to increase the precision of the proposed method. Through the experiments, the detection precision of the proposed method is 99.16%. This result has shown that the planned method has considerably higher efficiency and precision than other existing methods.