Predictive Analytics of Image Descriptors for Students Biometric Authentication System

Simple item page
dc.contributor.authorBabatunde Taiwo OLOMOLA
dc.date.accessioned2024-07-01T11:29:24Z
dc.date.available2024-07-01T11:29:24Z
dc.date.issued2023-12
dc.description.abstractNowadays, our education as well as other sectors is in a verge where accuracy of its authentication process is vital so as to close door tightly against impostors and impersonators. This thesis therefore improved the effectiveness and accuracy of biometrics based authentication models already in use for students’ attendance. The authentication framework is a five-phase biometric-based student attendance verification system that combined iris and fingerprint recognition attributes for the purpose of training deep learning models. The first phase entails the image acquisition. The acquired image inputs were then subjected to the feature extraction phase where attributes were extracted from the image inputs in form of numeric image descriptors. Data resampling were done in order to ensure a balanced training set for the machine learning-based study, which were consequently deployed for the deep learning phase after which performance evaluation was carried out in phase five. Three learner algorithms of Decision Tree (DT), Support Vector Machine (SVM), and Sequential Minimal Optimization (SMO) were trained with numeric vectors extracted from both fingerprint and iris biometrics of students for a student attendance authentication system. The numeric vectors were extracted using the SqueezeNet, InceptionV3, VG16, VG19, and Painters image embedders to return five distinct databases. The performances of the three base learners’ algorithms were evaluated alongside the performance of a Vote ensemble model after the five databases are subjected to a synthetic minority oversampling. Experimental results returned the Vote ensemble as the best model for student authentication which is followed by the SMO. The F1 score of Vote ensemble outperforms other models across the five datasets, with accuracy score as high as 0.999. The synthetic minority oversampling of the training sets further improved the performance of the models through data resampling. Consequently, Vote ensemble machine learning is better deployed for student authentication systems with any of the five image embedders. Keywords: Ensemble Machine, Information Security, Biometric Authentication, Biometric Recognition, Database Management System Word Count: 273
dc.identifier.citationKate Turabian
dc.identifier.otherM.Sc
dc.identifier.urihttps://repository.lcu.edu.ng/handle/123456789/631
dc.language.isoen
dc.publisherLead City University
dc.relation.ispartofseriesM.Sc
dc.subjectEnsemble Machine
dc.subjectInformation Security
dc.subjectBiometric Authentication
dc.subjectBiometric Recognition
dc.subjectDatabase Management System
dc.titlePredictive Analytics of Image Descriptors for Students Biometric Authentication System
dc.typeThesis

Files

Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
Olomola Babatunde Taiwo Full Thesis.pdf
Size:
19.48 MB
Format:
Adobe Portable Document Format
Download
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed to upon submission
Description:
Download

Collections

Department of Computer Science;