Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


ThesisManuscript-FirasAssaad-Final.pdf (2.7 MB)

ETD Abstract Container

Abstract Header

Biometric Multi-modal User Authentication System based on Ensemble Classifier

Assaad, Firas Souhail
http://orcid.org/0000-0002-1046-5093
http://rave.ohiolink.edu/etdc/view?acc_num=toledo1418074931

Abstract Details

2014, Master of Science, University of Toledo, Engineering.
User authentication performed through the traditional method of authorizing based on username and password results in the weakest link between users and their authentication credentials. Stolen authentication credentials in the form of user id and password pairs result in countless large-scale and detrimental security breaches across many segments of the social and economic groups. Authentication based on biometric features offer a much stronger link between the users and their credentials. In this project, we propose a multi-modal biometric authentication methodology to provide a more trusted authentication of the actual user. The system utilizes two biometric traits in its authorization procedure; these are face and voice recognition. During training, the face recognition subsystem is responsible for detecting the face in an image using the Viola-Jones algorithm, and implementing face preprocessing steps of eye detection and several geometric transformations to filter out unneeded details of the face. Using the Eigenfaces technique, the face recognition subsystem trains on those multiple images of a user being authorized for access and stores the resulting user-specific templates in the authorized users database. Similarly, during training, the voice recognition module acquires a voice sample from the user, extracts the voice features using the Mel-Frequency Cepstral Coefficients that are used to represent the “voiceprint” of the user, and then models those features using the Linde–Buzo–Gray algorithm. Each of the two modules, namely face recognition and voice recognition, following training performs as an independent base classifier within an ensemble design. The outputs from these two modules are fused together using score-level transformation to generate a final decision to either grant or deny access to a user who is in the process of authentication. The proposed algorithm has been trained and tested on Yale Extended, NIST FERET, and ELSDSR databases. System performance evaluation for authentication was tested in real time using a distributed framework that employed cellular phones or PDAs with cellular and WiFi connectivity, GSM networks, and a personal computer based server that is connected to the Internet. The ensemble classifier based authentication system performed as follows: accuracy at 99.22%, true positive at 99.15%, false positive at 0.71%, true negative at 99.28%, false negative at 0.84% and finally, precision at 99.24%. The system authentication response times were small enough to facilitate real-time implementation. We believe these results indicate that a biometric authentication system can reliably function as a supplementary to the existing password-based authentication procedures to result in a more trusted access control procedure.
Gursel Serpen (Advisor)
Jackson Carvalho (Committee Member)
Lawrence Thomas (Committee Member)
138 p.
Artificial Intelligence; Bioinformatics; Computer Engineering; Computer Science
algorithm; authentication; biometric; classifier; credentials; eigenfaces; multi-modal; recognition; score level; fusion; MFCC; Viola Jones; LBG; Mel Frequency Cepstral Coefficients; voice print; Linde Buzo Gray; ELSDSR; Yale Extended; NIST FERET; Server client;

Recommended Citations

Citations

  • Assaad, F. S. (2014). Biometric Multi-modal User Authentication System based on Ensemble Classifier [Master's thesis, University of Toledo]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=toledo1418074931

    APA Style (7th edition)

  • Assaad, Firas. Biometric Multi-modal User Authentication System based on Ensemble Classifier. 2014. University of Toledo, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=toledo1418074931.

    MLA Style (8th edition)

  • Assaad, Firas. "Biometric Multi-modal User Authentication System based on Ensemble Classifier." Master's thesis, University of Toledo, 2014. http://rave.ohiolink.edu/etdc/view?acc_num=toledo1418074931

    Chicago Manual of Style (17th edition)

toledo1418074931
2,750
© 2014, all rights reserved.
This open access ETD is published by University of Toledo and OhioLINK.
Release 3.2.12