Deteksi Pemalsuan Tanda Tangan Menggunakan Algoritma Convolutional Neural Network (CNN)

Abstract

A signature is a person's identity. This makes the existence of a signature necessary, usually taken from the name of a person or other style. Making signatures should not be changed because it will greatly affect the inauthenticity of identity, which is considered to be able to forge an important document and make transactions. Signature verification is mainly done manually, namely by comparing using the sense of sight, which still allows fraud to prevent signature forgery problems. It takes software that can perform original signature recognition to minimize fraud. This research built a signature recognition software that was done offline. The data used in this study were 144 signature images of lecturers at the computer science faculty, Universitas Indo Global Mandiri. The software implements a Deep Learning system by using the Convolutional Neural Network (CNN) algorithm. The CNN model is built with Python programming language and consists of a convolution, pooling, and fully connected layer. The research began with the preprocessing stage, followed by the construction of the CNN model and software testing. The input in this application is a signature image taken using a scanner. The output is in the form of the name of the signature owner and the accuracy of the match between the real and fake signatures. From the research that has been done, it got an accuracy value of 89.65% and a loss of 0.6018