Alliance International Conference on Artificial Intelligence and Machine Learning (AICAAM), April 2019 192 Siamese Triple Ranking Convolution Network in Signature Forgery Detection Ojaswini Chhabra * , Souradip Chakraborty * * Data & Analytics, Walmart Labs, Bangalore Abstract- Identifying a credible signature match based on a base signature of a person is an age-old problem. Despite recent automation and advances in this field using image recognition, a lot remains to be explored. In this paper, we develop an intelligent framework which can automatically detect a forged signature even if it is highly skilled, based on the developed feature embeddings and the corresponding algorithm. Siamese Triplet Convolution Neural Network is used to generate the feature embeddings for the signature images followed by a generalized Logistic Regression model to detect forgery. On the widely used SigComp dataset, our system achieves an accuracy of 96% in detecting forged signatures. Once the model is trained, it requires just one base image to determine whether another signature image is genuine or fraudulent with one shot learning. This algorithmic framework can be used in multiple commercial settings. One such example is testing customer or employee signatures on documents against a corresponding base signature saved beforehand. Index Terms- Active and real-time vision, Fraud detection, Off-line signature recognition, Triplet loss. I. INTRODUCTION ignatures are widely relied upon for identity verification by business, financial organizations and governments to authorize transactions and documents. Accurate signature verification is imperative since forgery and fraud can cost organizations money, time, and their reputation. In the last few years, a lot of progress has been made in the field of automating signature forgery detection using machine learning and image recognition-based concepts. Signature forgery can be broadly of two types: • Blind Forgery: Where the forger has no idea what the signature to be forged looks like. This is easy to detect by machine because it is usually not very close to the appearance of a genuine signature. • Skilled Forgery: Either simulation or tracing, in which the forger has a sample of the signature to be forged. In this case, detecting fraud requires more sophisticated tools to differentiate minute but critical details between genuine and forged signatures. In this paper, an automatic off-line signature verification and forgery detection system using image processing and Deep Convolutional Siamese networks is proposed wherein a deep triplet ranking network is used to calculate the image embeddings. This is coupled with generalized linear model architecture with logistic loss functions and cross validation to arrive at the final model to label images as authentic or forged. Training the model requires significant computation resources, but once the model is trained, it requires only one base image to determine whether another signature image is genuine or not with one shot learning. This process is instantaneous and can be carried out in real time. The main contribution of this paper is to enhance the robustness of the signature image embeddings using a FaceNet [1] based triplet network architecture with transfer learning using MobileNet CNN architecture [2]. The triplet loss S