A Bio-metric-based authentication and verification system for students in Makerere University

Abstract

Makerere University employs the method of examination permits in order to validate it’ s students and ensure overall completion of tuition payment. Although this method has kept the institution running for a long time now, it has major drawbacks. Forgery of examination permits and impersonation has long crept into our education system leaving a desperate need for a more efficient authentication and verification system in its wake. Also, relying on examination permits is absurd because paper is easily corruptible, damaged, lost or even forgotten, and all of these instances gravely inconvenience both students and administrators.This research work designed and implemented aBio-metric-based identity verification System to be used as a verification system to allow students of Makerere University into examination rooms. The system replaces the use of examination permits which the University has been using previously with use of individual fingerprints instead. On running their fingerprint through a biometrics fingerprint reader, the student’ s fingerprint iscaptured and the system uses that it to implement a principle of Machine Learning, the Convolutional Neural Network to identify the student to whom that fingerprint belongs from a database containing all student’ s information. This network was trained using a data set of fingerprints until it successfully “ learned” to identify a given fingerprint from a collection of other different fingerprints and return it’ s student number. On identifying the student number of that fingerprint, their corresponding financial and registration status for the paper they intend to sit is verified by the system, and if they are in the clear the arduino signals with a green light. If, however the student has not completed their tuition payments, or did not register for the examination that is being sat, it signals a red and white light respectively to alert the invigilators. Methodology used in the development of this system included Tensorflow for developing the Convolutional Neural Network to implement Machine Learning in this project. We also used C++ to program the LED lights on the arduino hardware to light differently according to a different outcome while the databases were run using MySQL. The system was tested to check for faults and errors. After training the neural network with a sample data set, it was tested to identify the corresponding student number of individual fingerprints and it successfully returned student numbers with an accuracy of up to 98.0%. Achievements and limitations faced during the course of this project are attached to this report, alongside with conclusions and recommended future works.