A two-stage forgery detection and localization framework based on feature classification and similarity metric

The data transfer in various forms, such as text, images, videos, etc., using the internet has become a routine exercise. Significant technological advancement has also been made by developing data editing software to make effective and efficient data transfer. On the one side, such advances provide substantial benefits such as easy and safe data transfer. Still, on the other side, human beings started the inappropriate use of these data editing software for their benefit. Surveillance videos and footage are considered the main sources of evidence for any crime in the present time. But, the 100% reliance on these videos is not feasible. In this paper, a two-stage inter-frame forgery detection technique has been proposed for HEVC-coded videos. The first stage detects the abnormal points based on compression domain features, and the second stage validates the abnormal points along with the forgery localization. Experimental results show that the proposed technique performs better with precision, recall, and F1-score of 0.9589, 0.9655, and 0.9622, respectively, at a low computational cost.