Nowadays, institutions such as banks, educational establishments, and government offices heavily rely on digital documents. Consequently, the issue of counterfeit paperwork has become increasingly significant. Forged identification cards, degrees, and official documents often result from simple photo editing software. Manually verifying each document is time consuming, allowing numerous scams to go unnoticed. Nonetheless, identifying forgeries remains more challenging than it ought to be. This research presents a technique for detecting fraudulent documents by integrating machine learning with image processing tools. By examining scanned documents closely, it pinpoints sections that appear modified—such as altered text, swapped images, and forged signatures. Prior to beginning the analysis, images undergo a series of cleanup procedures, followed by the automatic extraction of essential information. Software that recognizes letters and numbers from images scans the text. A specialized neural network aimed at identifying patterns in visuals aids in determining what seems out of place. Experiments conducted on different samples revealed high success rates in detecting altered files. Institutions managing official documents could benefit from this method of verifying authenticity.
Sarawade et al. (Thu,) studied this question.