Introduction

 
Individuals have always been identified by their behavioral and physiological traits. When dealing with this type of identification the most attention is towards biometrics. As Individuals entail features such as voice, iris, facial expressions, and fingerprints. Since there is an exponential rise in computation power and speed, theories and concepts developed years back have now become a reality.
Nowadays biometric systems are commonly utilized for information systems i.e. passwords or for tokenized systems like cards of identification for validations. Eg: drivers license, keys. Only biometrics are currently used for the mentioned systems as older technologies have become unreliable as information is easily leaked or lost. Due to this, there has been a substantial increase in the use of biometrics such as fingerprint sensors for validation of user login and online payments. So, when a user wants to authenticate himself through his fingerprint, the features are selected and then compared with the complete sets of fingerprints from a database.
When the database used for this process is big, a problem may arise when implementing for applications in real-time due to a large number of comparisons required increasing the computation power and time required.
Nowadays there are many fingerprint based systems that authenticate at high speeds that is because of the detailed clarity of the input provided to these algorithm. When there is a sample data provided to them with less detailed features it mostly leads to misclassification. We heavily rely on powerful sensors to provide us with great input. However, in this paper we would be discussing how we can achieve similar accuracy levels by utilizing secure verification methods to similarly classify input with substantial noise.This research will show the performance of using enhanced verification platforms along side deep learning structures over traditional systems.
This research will show the benefits of using enhanced verification platforms along side deep learning structures over traditional systems.
This research will detail the state-of-the-art, provides a description of the database used in this work, shows the architecture of the adopted networks, the parameters for training, and the evaluation protocol. Finally, the results will be presented and discussed, and the conclusions are drawn.