The ubiquity of smartphones has transformed them into primary repositories of sensitive data; however, traditional one-time authentication mechanisms create a critical trust gap by failing to verify identity post-unlock. Our aim is to mitigate these vulnerabilities and align with the Zero Trust Architecture (ZTA) framework and philosophy of “never trust, always verify,” as formally defined by the National Institute of Standards and Technology (NIST) in Special Publication 800-207. This study introduces a robust continuous authentication (CA) framework leveraging multimodal behavioral biometrics. A dedicated application was developed to synchronously capture touch, sliding, and inertial sensor telemetry. For feature modeling, a heterogeneous deep learning pipeline was employed to capture modality-specific characteristics, utilizing Convolutional Neural Networks (CNNs) for sensor data, Long Short-Term Memory (LSTM) networks for curvilinear sliding, and Gated Recurrent Units (GRUs) for discrete touch. To resolve performance degradation caused by class imbalance in Zero Trust environments, a Grid Search Optimization (GSO) strategy was applied to optimize a weighted voting ensemble, identifying the global optimum for decision thresholds and modality weights. Empirical validation on a dataset of 35,519 samples from 15 subjects demonstrates that the optimized ensemble achieves a peak accuracy of 99.23%. Sensor kinematics emerged as the primary biometric signature, followed by touch and sliding features. This framework enables high-precision, non-intrusive continuous verification, bridging the critical security gap in contemporary mobile architectures.

Feature-level fusion in multi-algorithmic and multimodal biometric systems produces very high-dimensional feature spaces that significantly degrade classification performance, increase computational and memory requirements, and reduce system scalability. This paper presents a binary Cuckoo Search (CS)–based feature selection framework designed to identify optimal feature subsets from fused biometric feature vectors while maintaining high recognition accuracy. The proposed approach encodes each feature subset as a binary nest, employs Lévy-flight–driven global exploration, implements local random walk mechanisms, and optimizes a wrapper-based fitness function combining recognition accuracy and feature subset size. Experimental evaluations on multi-algorithmic fingerprint, iris, and palmprint systems using public databases (CASIA V1.0, IITD V1.0, FVC, and others) demonstrate that CS-based feature selection substantially improves recognition rates compared to Principal Component Analysis (PCA) while achieving 80–91% feature-space reduction across various multi-algorithmic and multimodal configurations. Results show recognition accuracy improvements from 80–85% (PCA baseline) to 89–98% (proposed CS-FS) with Euclidean distance matching, and up to 99% accuracy with supervised classifiers. The method outperforms traditional dimensionality reduction techniques and exhibits competitive or superior performance relative to other evolutionary feature selection approaches such as Genetic Algorithms and Particle Swarm Optimization.

Reliable and robust personal authentication technologies have become indispensable in modern digital and physical security infrastructures. Traditional unimodal biometric systems—using a single biometric trait—often suffer from noise, spoofing vulnerabilities, and intra-class variability. To overcome these limitations, multimodal biometric systems fuse evidence from multiple biometric sources. However, feature-level fusion, despite yielding richer discriminatory information, produces high-dimensional feature spaces that demand efficient dimensionality reduction or feature selection. This review presents a consolidated analysis of three optimization-driven multimodal biometric recognition systems: Particle Swarm Optimization (PSO) for fingerprint–palmprint fusion, Genetic Algorithm (GA) for iris–fingerprint fusion, and Artificial Bee Colony (ABC) optimization for iris–palmprint fusion. We critically examine preprocessing techniques, feature extraction schemes, fusion strategies, dimensionality-reduction approaches, classifier performance, and comparative advantages. The review highlights trends, challenges, and future research directions in optimization-enhanced multimodal biometrics.

A Deep Learning-Based Multimodal Biometric Authentication Framework Using Fingerprint and Iris with Score-Level Fusion

In real-world applications, the reliability of biometric recognition systems that are based on 2D modalities is typically reduced due to limitations such as sensitivity to changes in illumination, facial expressions, and occlusion among other things. To overcome these problems, this research offers a multimodal biometric model that incorporates data from 3D face and 3D ear to achieve reliable identity recognition. The 3D biometrics offer more comprehensive structural information than the 2D characteristics, and they are more resistant to the effects of environmental changes. These 3D features are then used to recognize and secure storage of multimodal biometrics. Initially pre-processing steps, including cropping, normalization, hole filling, and spike removal are applied on 3d biometrics. After that, feature extraction is performed using the PointNet + + model, which is a network based on Convolutional Neural Networks (CNNs) that processes point clouds directly. We used the Face Recognition Grand Challenge (FRGC) database for 3D images of faces and the University of Notre Dame (UND) Collection G database for 3D images of ears for the tests. Our tests show that the PointNet + + model is accurate 99% of the time for 3D face recognition and 98% of the time for 3D ear recognition. With its 3D point cloud optimization and resilient architecture, the PointNet + + model achieves high accuracy for 3D face and 3D ear by learning multi-scale features that capture both local and global information.

As quantum computing continues to advance, it threatens the long-term protection of traditional cryptographic methods, especially in biometric authentication systems where it is important to protect sensitive data. To overcome this challenge, we present a comprehensive, privacy-preserving framework for multimodal biometric authentication that can easily integrate any two binary-encoded modalities through feature-level fusion, ensuring that all sensitive information remains encrypted under a CKKS-based homomorphic encryption scheme resistant to both classical and quantum-enabled attacks. To demonstrate its versatility and effectiveness, we apply this framework to the retinal vascular patterns and palm vein features, which are inherently spoof-resistant and particularly well suited to high-security applications. This method not only ensures the secrecy of the combined biometric sample, but also enables the complete assessment of recognition performance and resilience against adversarial attacks. The results show that our approach provides protection against threats such as data leakage and replay attacks while maintaining high recognition performance and operational efficiency. These findings demonstrate the feasibility of integrating multimodal biometrics with post-quantum cryptography, giving a strong, privacy-oriented authentication solution suitable for mission-critical applications in the post-quantum era.

Objectives: To propose a Deep Neural Spectral Subtraction Centroid Mel Frequency Spatial-Temporal (DN-SCMFST) framework for robust multimodal biometric authentication (face and voice). Method: The system uses the Speaking Faces dataset (142 subjects, 13,000 samples). Preprocessing is performed with a Kushner–Stratonovich filter and spectral subtraction. Features are extracted via spectral centroid and spatio-temporal descriptors, followed by score-level fusion. Findings: DN-SCMFST achieved accuracy improvements of 5–18%, PSNR gains of 18–28%, FAR reduction of 20–37%, FRR reduction of 18–28%, and recognition time reduction of 15–27% compared with CNN+RNN and DL-MBA models. Novelty: Integrates deep neural spectral subtraction with centroid-based mel frequency and temporal-spatial fusion, offering resilience against noise and improved real-time recognition efficiency. Keywords: Multimodal biometric authentication, Deep neural network, Spectral subtraction, Spectral centroid, Mel frequency, Score-level fusion, Kushner–Stratonovich

Recent developments in biometric recognition analysis include various imaginative intelligence concepts for identifying multimodal friction rides based on feature descriptions to solve the complexities in security and person identification applications. For information security, the Research uses multimodal screening analyses from finger veins, finger knuckles, iris, palm prints, and fingerprints. Most traditional methodologies only concentrate on object region regions, which fails to identify the feature relation and entity variation. The problem arises from the multi-correlation feature dimension, which increases the non-relation image registration and structural properties in entity resolution. Due to higher degradation in entity resolution, finding ridges creates false occlusion, causing a higher dimensionality ratio, lower precision rate, recall rate, and more false positives to provide lower identification accuracy. To resolve this problem, we propose an artificial intelligence based on AI-powered Multimodal biometrics recognition using deep scalar feature engineering with Optimal Particle Swarm Intelligence (OPSI) feature selection with Hyper capsule Gated LenetCNN. Initially, the multimodal biometric dataset is pre-processed with an adaptive Gaussian mean filter to normalize the images, and the Iterative pattern Slice Fragment Clustering (IPSFC) is applied for entity markings. Then, a vector Pyramid Scene Parsing Network is used for segmentation. Then, to reduce the feature dimension, the Optimal Particle Swarm Intelligence (OPSI) feature selection is applied to minimize the non-relation features. Then, LSTM- Hyper Capsnet CNN is used to identify the biometric data. Our methodology involves extracting pertinent biometric data features and utilizing deep learning algorithms to enhance identification accuracy. The proposed deep optimal feature engineering approach is introduced, allowing the model to prioritize and select features that significantly increase the identification accuracy while minimizing feature redundancy. The output result is evaluated, and improved accuracy, sensitivity, specificity, and F1 measure performance are achieved, significantly increasing the precision rate based on the proposed CNN classification.

Cyber coercion, where legitimate users are forced to perform actions under duress, poses a serious insider threat to modern organizations, especially to critical infrastructure. Traditional security controls and monitoring tools struggle to distinguish coerced actions from normal user actions. In this paper, we propose a cyber coercion detection system that analyzes a user’s activity using an integrated large language model (LLM) to evaluate contextual cues from user commands or actions and current policies and procedures. If the LLM indicates coercion, behavioral methods, such as keystroke dynamics and mouse usage patterns, and physiological signals such as heart rate are analyzed to detect stress or anomalies indicative of duress. Experimental results show that the LLM-assisted multimodal approach shows potential in detecting coercive activity with and without detected coercive communication, where multimodal biometrics assist the confidence of the LLM in cases in which it does not detect coercive communication. The proposed system may add a critical detection capability against coercion-based cyber-attacks, providing early warning signals that could inform defensive responses before damage occurs.

Biometric authentication is a cornerstone of modern security systems, yet concerns regarding privacy and data security persist. Cancellable biometrics offer a solution by transforming raw biometric data into non-invertible representations, ensuring security even in the event of a data breach. This study presents Multimodal Affine Cover-space Euler Transformation (MACET), a novel framework designed to enhance biometric template security while preserving authentication accuracy. The proposed approach is based on the hypothesis that Affine Cover Space transformation combined with Euler’s form can generate irreversible templates for multimodal biometrics, specifically fingerprint and iris data, without compromising recognition performance. The methodology involves feature extraction, inverse matrix computation, affine transformation, and Euler-based augmentation, ensuring robust and secure biometric template generation. Experimental results, conducted on a dataset of 450 biometric samples, demonstrate the effectiveness of MACET in improving authentication performance. The system achieves an Equal Error Rate (EER) of 0.0046 and an Area Under the ROC Curve (AROC) of 0.9886, indicating high accuracy. Additionally, the method significantly reduces storage memory size to 1.37 KB per template while maintaining an average execution time of 10.89 seconds. Robustness analysis against spoofing attacks confirms the system's ability to resist unauthorized access, ensuring strong security and privacy protection. These findings establish MACET as a highly secure, computationally efficient, and privacy-preserving biometric authentication framework, suitable for real-world applications. Future research could extend this approach to additional biometric modalities and large-scale authentication systems.

Biometrics plays a crucial role in mitigating threats such as theft, duplication, and cracking by offering more secure verification methods. To enhance system reliability, researchers are increasingly focusing on multimodal biometrics that integrate facial recognition and fingerprint identification. The objective is to design a biometric verification system that leverages deep learning to automatically extract and analyze features from fingerprints, videos, and facial images. This system employs image scaling and data augmentation during preprocessing to preserve information and reduce computational time. To strengthen resistance against software attacks and varying poses, dynamic fusion techniques applied to hand-surface features are incorporated. Furthermore, multi-scale single-shot face detectors enable efficient face detection in unconstrained videos, while memory-efficient deep neural networks (DNNs) ensure optimal resource utilization. The study applies advanced approaches such as Transfer Learning and Hyperparameter Optimization algorithms, including Keras Tuner (Random Search), Genetic-CNN, Teaching Learning Based Optimization (TLBO), and Grey Wolf Optimizer (GWO). Findings demonstrate that models integrated with hyperparameter optimization significantly outperform those without optimization. For facial recognition, CNN-GA achieved an impressive classification accuracy of 99.75%, while in fingerprint recognition, Keras Tuner recorded a peak accuracy of 99.09%. These outcomes highlight the effectiveness of combining deep learning with optimization strategies in building robust multimodal biometric systems. By integrating efficient preprocessing, adaptive algorithms, and optimized architectures, the proposed framework not only enhances accuracy but also ensures resilience against diverse attack vectors, positioning multimodal biometrics as a key solution for future secure authentication technologies.

Hand-based multimodal biometrics have attracted significant attention due to their high security and performance. However, existing methods fail to adequately decouple various hand biometric traits, limiting the extraction of unique features. Moreover, effective feature extraction for multiple hand traits remains a challenge. To address these issues, we propose a novel method for the precise decoupling of hand multimodal features called 'Normalized-Full-Palmar-Hand' and construct an authentication system based on this method. First, we propose HSANet, which accurately segments various hand regions with diverse backgrounds based on low-level details and high-level semantic information. Next, we establish two hand multimodal biometric databases with HSANet: SCUT Normalized-Full-Palmar-Hand Database Version 1 (SCUT_NFPH_v1) and Version 2 (SCUT_NFPH_v2). These databases include full hand images, semantic masks, and images of various hand biometric traits obtained from the same individual at the same scale, totaling 157,500 images. Third, we propose the Full Palmar Hand Authentication Network framework (FPHandNet) to extract unique features of multiple hand biometric traits. Finally, extensive experimental results, performed via the publicly available CASIA, IITD, COEP databases, and our proposed databases, validate the effectiveness of our methods.

Multimodal biometrics technology has garnered attention recently for its ability to address inherent limitations found in single biometric modalities and to enhance overall recognition rates. A typical biometric recognition system comprises sensing, feature extraction, and matching modules. The system’s robustness heavily relies on its capability to effectively extract pertinent information from individual biometric traits. This study introduces a novel feature extraction technique tailored for a multimodal biometric system utilizing electrocardiogram (ECG) and iris traits. The ECG helps to incorporate the liveliness related information and Iris helps to produce the unique pattern for each individual. Therefore, this work presents a multimodal authentication system where data pre-processing is performed on image and ECG data where noise removal and quality enhancement tasks are performed. Later, feature extraction is carried out for ECG signals by estimating the Heart rate variability feature analysis in time and frequency domain. Finally, the ensemble of convolution neural network (CNN) and DeepResNet models are used to perform the classification. The overall accuracy is reported as 0.8900, 0.8400, 0.7900, 0.8932, 0.87, and 0.97 by using convolutional neural network-long short-term memory (CNN-LSTM), support vector machine (SVM), random forest (RF), CNN, decision tree (DT), and proposed MBANet approach respectively.

ObjectiveThis study aims to enhance cybersecurity by implementing a robust biometric-based authentication approach. A Multimodal Biometric System (MBS) is proposed, utilizing feature-level fusion of human facial (physiological) and speech (behavioral) features to improve security, accuracy, and user convenience. The system addresses the limitations of traditional authentication methods, including unimodal biometrics and password-based security.BackgroundIn the modern digital landscape, human-computer interaction and digital platforms play a crucial role in daily life. With billions of users engaging in social media, financial transactions, and e-commerce, the demand for secure authentication mechanisms has intensified. However, the increasing sophistication of cyber threats poses significant risks, undermining trust, security, and confidence in digital systems.Method: The proposed MBS incorporates improved proposed techniques for feature extraction, feature level fusion strategies and an ensemble classification model combining Bi-LSTM and DCNN. To optimize performance, the system is enhanced using an improved bio-inspired Manta Ray Foraging Optimization (MRFO) algorithm.ResultsThe system's performance was evaluated using two publicly available Voxceleb1 and VidTIMIT datasets, achieving accuracy rates of 98.23% and 97.92%, with Equal Error Rates (EERs) of 3.23% and 3.62%, respectively.ConclusionThe proposed approach outperforms conventional optimization techniques and existing state-of-the-art MBS. As a contactless and non-intrusive authentication system, it enables seamless data acquisition through devices equipped with cameras and microphones, such as smartphones, ensuring real-time processing of biometric modalities.Application: This contactless MBS presents a viable solution for secure and hygienic authentication in applications requiring high cyber resilience, including banking, e-commerce and other digital security domains.Precis/Table of Contents: This research enhances cybersecurity by proposing a Multimodal Biometric System (MBS) that integrates feature-level fusion of facial (physiological) and speech (behavioral) traits. The approach improves security, accuracy, and user convenience while addressing hygiene concerns. It overcomes the limitations of traditional authentication methods, including unimodal biometrics and password-based security vulnerabilities.

Multimodal Biometric System (MBS) is an advanced progression of conventional biometric authentication system, which employ multiple biometric traits to enhance security. However, despite their advantages, these systems are vulnerable to presentation attacks, where adversaries use photos, replay videos or voice recordings to deceive the authentication process. Therefore, this paper proposes a challenge-response based approach using texture-based facial features and multidomain speech features. The challenge-response approach requires the user to utter a random word. Next, the system detects the user’s facial features (eye and mouth motion) and recognized speech text to confirm whether the authentication request originates from a legitimate user or an imposter. The feature-level fusion via concatenation method is used to combine these image-audio features, to reduce the overlap within the feature spaces and data dimensionality. The fused feature vector is then fed into the deep learning driven ensemble classifier CNN-BiLSTM to train and test the fused samples for user authentication. The performance evaluation is carried out using a self-built database with 55 users, achieving 96.81% accuracy, 98.20% precision and an Equal Error Rate (EER) of 3.37%. Moreover, the proposed approach surpasses different cutting-edge MBS, deep learning classifiers and image-audio fusion techniques on various performance metrics. Thus, the results underscore the effectiveness of the deep learning-based MBS in ensuring user authentication and spoof detection, demonstrating its considerable potential in bolstering the security of biometric systems against intricate presentation attacks.

This study introduces a blockchain-based electronic voting system enhanced with fingerprint and facial recognition to address persistent challenges such as fraud, impersonation, and lack of transparency in electoral processes. Centered on Nigeria's electoral context and supported by data from the 2024 Edo State gubernatorial election, the system utilizes blockchain’s decentralized and tamper-proof ledger for secure vote recording, while multimodal biometrics ensure real-time, accurate voter authentication. Using SWOT analysis and IBM SPSS for statistical evaluation, the framework demonstrates improved security, transparency, and operational efficiency compared to conventional systems. Despite existing challenges related to infrastructure, biometric variability, and data privacy, the model presents a scalable, future-ready solution that can reinforce democratic legitimacy and restore public trust in elections. The findings contribute to advancing research in e-governance, secure computing, and electoral reform, advocating for technology-driven policy changes in emerging democracies.

Adaptive Parameter based Mantis Search Algorithm and Regularized Dropout with Gated Recurrent Unit for Multimodal Biometric Recognition

The widespread adoption of e-health systems raises critical concerns regarding data privacy and network security. Ensuring secure and reliable data sharing between patients and healthcare professionals remains a significant challenge. To address this, we propose a novel anonymous authentication scheme tailored for e-health environments, integrating zero-knowledge proof (ZKP) with multimodal biometrics. Our key contributions are as follows: (1) applying the Pedersen vector commitment algorithm to construct a biometric-based ZKP scheme, thereby ensuring enhanced security and privacy-preserving authentication; (2) utilizing multimodal cancelable biometrics generate (MCBG) technology, integrating fingerprint, face, and iris modalities to strengthen the security of the verification process; and (3) providing a detailed security analysis that demonstrates our scheme meets essential security requirements, including anonymity, authenticity, unlinkability, forward security, and resistance to replay attacks. Experimental results demonstrate stable proving and verification time of approximately 78 ms and 140 ms, respectively, regardless of the proof range, validating its efficiency and practicality for secure authentication in e-health systems.

The widespread adoption of e-health systems raises critical concerns regarding data privacy and network security. Ensuring secure and reliable data sharing between patients and healthcare professionals remains a significant challenge. To address this, we propose a novel anonymous authentication scheme tailored for e-health environments, integrating zero-knowledge proof (ZKP) with multimodal biometrics. Our key contributions are as follows: (1) applying the Pedersen vector commitment algorithm to construct a biometric-based ZKP scheme, thereby ensuring enhanced security and privacy-preserving authentication; (2) utilizing multimodal cancelable biometrics generate (MCBG) technology, integrating fingerprint, face, and iris modalities to strengthen the security of the verification process; and (3) providing a detailed security analysis that demonstrates our scheme meets essential security requirements, including anonymity, authenticity, unlinkability, forward security, and resistance to replay attacks. Experimental results demonstrate stable proving and verification time of approximately 78 ms and 140 ms, respectively, regardless of the proof range, validating its efficiency and practicality for secure authentication in e-health systems.

FaciaVox: A diverse multimodal biometric dataset of facial images and voice recordings.