Document Forgery Detection

Abstract

Document forgery is an increasing problem for both private companies and public administrations. It can be said to represent the loss of time and resources. There are many classical solutions to these problems such as the detection of an integrated security pattern. In such cases, it is important that we resort to forensic techniques for the detection. The idea behind using these forensic techniques can also be implemented using artificial intelligence/machine learning which can be of lower cost and can provide the same or better results. The experimental result shows that multiple models have strong detection capability to detect multiple forgeries. In this paper, we have developed a different approach to detecting forgery in a document. The forgery we detect can be classified as hand-written signature forgery and copy-move forgery of any photo, text, or signature. We have developed a novel approach using capsule layers to detect a forgery in handwritten signatures. We also use ELA (Error Level Analysis) to detect any error in the compression levels of the image.