Ashok Kumar Yadav, et. al. International Journal of Engineering Research and Applications www.ijera.com ISSN: 2248-9622, Vol. 11, Issue 2, (Series-V) February 2021, pp. 23-29 www.ijera.com DOI: 10.9790/9622-1102052329 23 | Page Fingerprint Authentication Using CNN for Minutiae based Authentication Ashok Kumar Yadav, DGM ECIL and Prof. T. Srinivasulu, K U Warangal ABSTRACT There has been an increase in security concerns regarding fingerprint based authentication. This problem arises due to technological advancements in spoofing and hacking technologies. This has motivated the need for a more secure platform for identification. In this paper, we have used a deep Convolutional Neural Network as a pre-verification filter to filter out fake or spoof fingerprints. As deep learning allows the system to be more accurate at detecting and reducing false identification by training itself again and again with test samples, the proposed method improves the security and accuracy by multiple folds. The implementation of a novel secure fingerprint authentication platform that takes the optical image of a fingerprint as input. The given input is pre-verified using Google’s pre-trained inception model for deep learning applications, and then passed through a minutia-based algorithm for user authentication. Then, the results are compared with existing models. Keywords: Biometrics, deep learning, convolutional neural network, inception model, minutiae, fingerprint. -------------------------------------------------------------------------------------------------------------------------------------- Date of Submission: 13-02-2021 Date of Acceptance: 27-02-2021 -------------------------------------------------------------------------------------------------------------------------------------- I. INTRODUCTION We, human beings, can be recognised by our unique characteristics and traits. These traits can be physical ,behavioural or physiological in nature. At present, when we consider identification, we immediately associate it with biometrics. Identification, in general, is done by observing and identifying the unique characteristics of an individual. These characteristics can be features such as facial expressions, iris patterns, voice, signature , fingerprint vein and even fingerprints. We have formulated many methodologies for identifying individuals based on their unique features, but most of them are left as theories due to the lack of computational power and speed. However, in this day and age, we can convert these theories into practical uses, and can make them available for day-to-day use [10] . Previously, identification and validation of individuals were done by information-based systems, which make use of passwords or cards for authentication. As these methods are less reliable and less secure, the stored information can be easily hacked or lost. Therefore, there has been a need for more secure, complex, and unique identifiers for user authentication. Hence, biometric traits such as such as fingerprints, palm prints, face, iris, and ECG are found to be very useful for the recognition of individuals. Among the above-mentioned biometric traits, fingerprint-based authentication is the most popular one. Because of the above- mentioned risks, people have adopted fingerprint sensors for biometric authentication in many financial transaction platforms [10]. Fingerprints have different patterns that make each individual fingerprint unique. The patterns are made by a combination of several ridges and valleys. These ridges are moulded during our time in the womb, where several factors such as friction, maternal conditions, etc. affect their final shape and structure. These patterns develop all over the human body, including the palms, soles, and even toes. As so many conditions and factors play an important role in determining the final ridge and valleys pattern, we consider fingerprint patterns to be unique to each and every individual. Fingerprint authentication has become a norm in our day-to-day society that its vulnerability is never challenged. However, due to technology advancements, malevolent attempts which bypass security systems using fake fingerprints have been increased. Many systems today use algorithms that match the records in the database with the input RESEARCH ARTICLE OPEN ACCESS