Article Preview
TopIntroduction
The mobile revolution has resulted in the development of portable and convenient data manipulation and data transfer devices. Authentication is a well-known method of securing a user's access to services. Unlocking devices, authorizing transactions, and verifying users' signatures are all examples of authentication services (Wojtowicz & Joachimiak, 2016). For both service providers and users, the traditional authentication method prioritizes using a username and password (Czeskis et al., 2012). However, physical characteristics combined with one or more biometrics allow for the precise identification of an individual user (Casanova et al., 2021). Biometric recognition utilizes user characteristics such as hands, eyes, faces, mouth, fingerprints, and ears to identify users automatically. The unimodal biometric system's limitations include a lack of data, low accuracy, and vulnerability to attack.
Biometric authentication is defined as the process of automatically recognizing physiological characteristics to provide secure identification and verification solutions (Selwal et al., 2016). Five modules comprise the biometric system: 1) a multi-sensor system collects primary biometric data, 2) a preprocessor for the necessary feature extractor methods, 3) templates generate patterns for biometric extracts, 4) decision fusion reconciles disparate patterns, and 5) identifying and authorizing the results by matching the input and enrolled patterns (Yadav et al., 2011). The multimodal biometric system expands the number of biometric inputs from users to more than two, allowing for more reliable authentication. Multiple physical characteristics can be used to authenticate users. Multimodal biometrics integrates and improves user recognition with increased accuracy, security, reliability (Purohit & Ajmera, 2021), and fault tolerance (Khoo et al., 2018), as well as consolidating data. The multimodal biometric system requires multiple parallel systems to detect the same or distinct biometric features at various layers (Smith & Brooks, 2013). Human physiological biometrics have been studied practically in a variety of areas using multimodal biometric systems, including parallel feature fusion of periocular biometrics and touch (Casanova et al., 2021), unequal feature portions of faces and fingerprints (Yang et al., 2019), and novel techniques and algorithms for palm, fingerprint, and ear recognition (Purohit & Ajmera, 2021).
Pukdesree and Netinant (2016) proposed a biometric authentication conceptual framework and developed a cloud-based biometric authentication service; thus, the entire process of data input, processing, and output can express the multimodal biometric authentication layering architecture. The framework provided a layered design for the client, application, and cloud database. The actual authentication system is still being developed and evaluated. While biometric authentication has evolved to incorporate multiple physiological biometrics with the possibility of a single point of failure, the multimodal biometric authentication system built on this framework can be enhanced by adding a security access system to protect fraud detection services.
This paper proposes a design and implementation of multimodal biometric cloud services for authentication of web-based examinations to bridge the gap between face and voice recognition multimodal biometric applications. During the coronavirus (COVID-19) pandemic, students could verify their identities and access the exam scheduling system. The system is unburdened within acceptable responsiveness scores to exam schedules, and the system was replaced with more flexible and superior securities. A confusion matrix is used to compare nonmembers and members who have registered for the examination system. The system outcomes from this biometric authentication research can be used to define system requirements and development scopes, as well as improve the reliability and accuracy of practical web-based applications.