HANDWRITTEN SIGNATURES FORGERY DETECTION

Abstract: Signature plays a very import role in sectors like banking, finance, Passport, Driving License, legal documentation etc. Signature varies from person to person and may be unique each time. Some time signatures may seem similar if the people have same name. But the features may still vary. Now a days there are problems like identity theft, fake ids, hacking etc. To reduce such type of issue, this project is focused on developing a system to detect such theft and to know and verify if the signature is real or fake, from the data sets using CNN and deep learning.The reason for using CNN and deep learning is that, the signature can vary with change in personalities and behaviour. With deep learning we can train the data sets and increase the accuracy of the detection. The Signatures can be hand written or signed online, depending on the type of signature the process takes place. Here we are referring to few papers which implement the project using both online and offline methods based on deep learning models. Using these we can try and achieve a better accuracy

Keywords: Signature, CNN, Forgery, Authentication, Deep learning