A. UNIMODAL BIOMETRICSThe unimodal biometric systems rely on the evidenceof a single source of information for authentication ofperson. Though these unimodal biometric systems havemany advantages, it has to face with variety problems likeNoisy data,Intra class variation,Interclass similarities,Nonuniversality,Spoofing etc.
B. TYPES OF MULTIMODAL SYSTEMSDepending on the traits, sensors and feature sets manydifferent types of multimodal systems are there. https://codeshoppy.com/php-projects-titles-topics.html These include
1) Single biometric trait, multiple sensors:Multiplesensors are used to record the same biometric characteristic.The raw data taken from different sensors can then becombined at the feature level or matcher score level toimprove the performance of the system.2) Multiple biometrics:Multiple biometric traits such asfingerprints and face can be combined.Different sensors areused for each biometric characteristic. The interdependencyof the traits ensures a significant improvement in theperformance of the system.3) Multiple units, single biometric traits:Two or morefingers of a single user can be used as a biometric trait. Itis inexpensive way of improving system performance, as itdoesnt require multiple sensors or incorporating additionalfeature extraction or matching modules. Iris can also beincluded in this category.4) Multiple snapshots of single biometric:In this morethan one instance of the same biometric is used for therecognition. For e.g. multiple impressions of the same fingeror multiple samples of the voice.5) Multiple matching algorithms for the same biometric:In it different methods can be applied to feature extractionand matching of the biometric characteristic.
C. FUSION LEVELS IN MULTIMODAL BIOMETRICSThere are three fusion levels in multimodal biometrics:feature level fusion, matching score level fusion and decisionlevel fusion respectively. The three levels of fusion aredescribed as follows:1) FEATURE LEVEL FUSION:In the feature levelfusion, features from different biometric traits are initiallyprocessed and the feature vectors are obtained are extractedand combined to form a composite feature vector. This isthen combined to form a feature vector that is used forclassification.2) MATCHING SCORE LEVEL FUSION:In matchingscore level fusion, individual matching score is found basedon various biometric traits and these matching scores aregathered to make the classification.3) DECISION LEVEL FUSION:In decision level fusion,each biometric traits are captured and features are extractedfrom the captured traits.The final decision of accept or rejectbased on the combination of the outcomes from differentbiometric modalities.
D. MATCHING ALGORITHMSBased on the pattern of the matching algorithm, thematching speed can vary. In a biometric recognition system,the individuality corresponding to the probe is clasicallydetermined by matching it against the templates of allindividualities in the gallery.
E. FINGERPRINT MATCHING TECHNIQUESFor accurate personal identification,considering all the cur-rently used biometric techniques, fingerprint authenticationsystem is the widely used and appropriate.The existing popularfingerprint matching techniques can be broadly classified intothree categories depending on the types of features used:1) Minutiae-based:2) Correlation-based:3) Euclidean distance-based:III. PROPOSEDSYSTEMDESIGNIn the proposed system, the online banking systemensures robust and secure authentication mechanism by usingthe multimodal biometrics.Multimodal system includingFingerprint and face are used for the login process. As theftcan occur at any point of transaction process, fingerprintauthentication is again done during transaction process.Efficient encryption and decryption methods are used forproviding the security of data transmitted and storing the datain the database. Thus the proposed system ensures improvedsecurity in online banking by using the multimodal biometricsystem.Figure 1.High level designFigure 1 describes the overall scenario in the proposedsystem.The planned system consists of a client system which isthe user doing the online transaction. The bank server enclosesthe database with which the details has to be compared. Theuser can login with the user id , and recognising self withfingerprint and face . These details are compared with thedatabase in the server. Once the login is successful, the usercan make the necessary transaction by authenticating with thefingerprint once again. The details are again compared withthe server.The proposed system uses a multimodal biometric system.Itconsists of two main modules namely,A. Enrolment moduleHere, the user has to register at the bank with the necessarydetails . This includes the biometric traits as well as information needed for the authentication.B. Authentication moduleHere, the user has to authenticate him/herself using themulti biometric traits used for the login process and unimodalbiometric , used for transaction process.The Authenticationmodule consist of two main processes.1) Login Process:Here, the user has to login using theuser id followed by the recognition of face and fingerprintfor authentication .Once the user login to the system the usercan only view the account details.2) Transaction Process:Here, the user has to againauthenticate him/her self using the fingerprint authentication.Only when the user authenticate with the fingerprint details,the transaction can be done.The authentication mechanism includes the processes atboth the client and server side. The client side process includecapturing the finger and face image ,followed by featureextraction and fusion of the feature extracted,encrypting theEuclidean distance calculated and sending it to the server.Thisis depicted in Figure 2.Figure 2.Client sideFigure 3 illustrates the server side process. The serverside process include, decrypting the encrypted data, andcomparing the stored data in the database.