Abstract This paper presents a secure multimodal biometric authentication framework based on cyclically coupled fractional-order hyperchaotic systems. The features of the face, iris and thumbprint are extracted using classical image descriptors and embedded as initial conditions of coupled hyperchaotic oscillators, transforming biometric authentication into a dynamic process driven by synchronization. The cyclic coupling topology enforces mutual dependency among biometric modalities, ensuring reliable authentication only when all traits originate from the same individual. Numerical analyses based on bifurcation diagrams and Lyapunov exponent spectra confirm the hyperchaotic behavior of the proposed system. Synchronization error analysis demonstrates robust convergence under appropriate coupling strengths, with fractional-order dynamics providing improved stability and faster convergence compared to integer-order counterparts. In addition, chaotic encryption produces noise-like, visually unrecognizable biometric templates, ensuring non-invertibility and template revocability. The proposed framework offers a secure and flexible solution for next-generation multimodal biometric authentication systems.

The quick development of digital communication platforms has opened up new areas where people’s behavior is profoundly and frequently shaped by online anonymity. This paper looks at the psychological understanding of online anonymity and how it can lead to harmful behaviors like hate speech, harassment, cyberbullying, and flaming. In order to provide a thorough understanding of why people behave differently when protected by anonymity in digital environments, the study shows findings from theoretical literature, drawing on important theoretical frameworks. The research paper shows three main factors of toxic online disinhibition: anonymity, invisibility and lack of eye-contact. The long-held belief that toxic behavior is solely motivated by anonymity is challenged by experimental findings showing that the single biggest cause of negative disinhibition is absence of eye-contact. The study also shows how cyberbullying appears in online forums, how anonymity is adversely correlated with aggressive attacks, and how algorithmic amplification and social media platform design amplify these effects. The study recognizes the dual nature of anonymity in addition to the risks: although it encourages toxic behavior, it also decreases barriers to self-disclosure, assists vulnerable people seeking mental health support, and creates positive anonymous networks. The effects of toxic online behavior are examined at the individual level, such as despair, anxiety, and suicidal thoughts among victims, as well as at the communal and societal levels, such as the dissemination of false information and heightened social division. AI-powered content filtering, digital literacy instruction, identity verification systems, and platform design reform are some of the methods to lessen toxic behavior that are covered.

Face anti-spoofing is an essential component in modern biometric authentication systems, ensuring that recognition technologies are not deceived by fraudulent attempts such as printed photographs, video replays, or 3D masks. This project proposes a Data Fusion-Based Two Stage Cascade Framework that integrates multiple modalities—RGB (Red, Green, Blue), Depth, and Infrared (IR)—to improve robustness and accuracy in detecting spoofing. In the first stage, deep learning models like 3D Convolutional Neural Networks (3D CNNs), CNN LSTM (Convolutional Neural Network with Long Short-Term Memory), and attention mechanisms are applied for feature extraction. In the second stage, their outputs are combined through decision fusion in a multi-stream network. The framework is evaluated on benchmark datasets like CASIA-SURF and Replay-Attack. Results show significant improvements over traditional single-modality systems, making the proposed framework suitable for real-world applications in banking, airport security, and access control systems.

Systemic data bias constitutes a major source of failure in real-world AI systems and represents a regulatory challenge that remains insufficiently addressed by existing legal frameworks, including the EU Artificial Intelligence Act. Although the AI Act introduces a comprehensive risk-based regulatory regime, it does not adequately capture how bias originates, propagates, and manifests across the AI lifecycle. This paper examines systemic data bias through a legal-technical lifecycle analysis that maps recurring bias mechanisms, from data collection and annotation to model training, evaluation, and deployment, to the regulatory control points established under the EU AI Act. Drawing on cross-sectoral examples from employment screening, credit scoring, healthcare risk prediction, biometric identification, and autonomous systems, the analysis demonstrates how technical bias mechanisms translate into systemic governance and accountability challenges. The findings reveal persistent regulatory gaps, including limited auditability of training datasets, the absence of mandatory fairness metrics, insufficient transparency regarding model behavior, and weak mechanisms for post-deployment monitoring and accountability. These results highlight a structural misalignment between lifecycle-based bias dynamics and the Act’s category-driven compliance framework. The paper argues that addressing systemic bias requires a governance approach that integrates technical bias mitigation with legal oversight across the full AI lifecycle rather than relying primarily on post hoc regulatory controls.

Smartphones have become indispensable in daily life, serving as hubs for communication, financial transactions, and data storage. However, their ubiquity and functionality make them prime targets for cyberattacks. This study delves into the development and enhancement of hacking detection mechanisms tailored to smartphones. It reviews existing techniques, including behavioural analytics, anomaly detection, machine learning, and intrusion detection systems, while identifying gaps in their effectiveness. A comprehensive framework for real-time detection is proposed, integrating multi-layered security measures, data encryption, and AI-driven analysis. Through simulations and case studies, the framework demonstrates improved detection rates and reduced false positives compared to traditional methods. This paper emphasizes the critical need for proactive, adaptive solutions to counter emerging threats in the rapidly evolving landscape of mobile security. The study explores the benefits of biometric identification over more conventional techniques like smart cards, PINs, and passwords since biometric qualities are hard to copy or steal. The study also looks at the advantages and disadvantages of the various kinds of biometric authentication systems. The promise of biometric authentication as a practical and safe means of identification in a range of applications is emphasized in the paper's conclusion.

Biometric authentication systems founded on face recognition are now embedded across a broad spectrum of real-world applications, making themhigh-value targets for spoofing attacks. Adversaries exploit artefacts ranging from two-dimensional printed photographs and replay video sequences to sculpted three-dimensional masks to deceive thesesystems. Countering such threats, face anti-spoofing (FAS)—also termed presentation attack detection (PAD)—has emerged as an indispensable safeguardwithin modern authentication pipelines. This paper presents a spatio-temporal deep learning framew ork that fuses a ResNet-50 spatial encoder with a GatedRecurrent Unit (GRU) temporal module to simultaneously capture liveness cues at both the texture and motion levels. Beyond the proposedsystem, a structured review of contemporary deep FAS methodologies is provided, covering pixel-w isesupervisory signals, domain-invarianttraining strategies, open-set evaluation protocols, and sensor-aware multi-modal architectures. Experimental results on a curated dataset of 1,250 samples yield aclassification accuracy of 96.2%, a false acceptancerate of 3.4%, and a false rejection rate of 3.8%, outperforming several recently published baseline methods. Key open challenges and prospective research directions are identified to guide further development of robust, deployment-ready FAS systems. Keywords—Face anti-spoofing; presentation attack detection; spatio-temporal deep learning; liveness detection; multi-modal generalization

The rapid development of smart technologies and digital infrastructure has transformed the way governments and organizations monitor and manage urban environments. Smart societies rely on advanced technologies such as artificial intelligence, Internet of Things devices, biometric identification systems, and large-scale data analytics to improve public services, enhance security, and optimize urban management. Digital surveillance systems have become a central component of smart city governance by enabling real time monitoring of public spaces, transportation systems, communication networks, and citizen activities. While these technologies provide numerous benefits in terms of public safety, crime prevention, and efficient resource management, they also raise significant concerns regarding privacy protection and civil liberties. Digital surveillance technologies collect and analyze vast amounts of personal and behavioral data, which may create risks related to unauthorized monitoring, data misuse, and excessive governmental control. Civil liberty advocates argue that widespread surveillance may undermine fundamental democratic principles such as freedom of expression, freedom of association, and the right to privacy. As smart societies continue to expand the use of surveillance technologies, it becomes increasingly important to understand how these systems influence public perceptions of privacy, governance transparency, and civil rights protection. This study analyzes the effect of digital surveillance systems on civil liberties within smart societies. The research develops a conceptual model that examines the relationships between digital surveillance intensity, perceived privacy risk, governance transparency, and protection of civil liberties. Data were collected from citizens, technology professionals, and policy analysts involved in smart city initiatives. Structural Equation Modeling using Smart Partial Least Squares was employed to test the relationships between constructs. The results indicate that increased digital surveillance intensity significantly raises perceived privacy risks among citizens. However, governance transparency and regulatory oversight mechanisms play a critical role in mitigating negative perceptions and protecting civil liberties. The study contributes to digital governance and technology policy research by providing empirical insights into the complex relationship between surveillance technologies and civil liberty protection in smart societies. The findings highlight the need for balanced governance frameworks that ensure security benefits while safeguarding individual rights and democratic values.

Multimodal biometrics are able to improve the accuracy and security of authentication by integrating more than one biometric characteristic, minimizing errors, and maximizing the resistance to attacks. The primary drawback of multimodal biometric verification is the complexity of the systems that are introduced by multiple sensors, more computing, and fusion issues. Multimodal feature extraction methods are inadequate in traditional feature extraction methods as they generate modality-specific, handcrafted representations which are not robust, compatible and discriminative enough to support effective feature-level fusion. Deep learning feature extractors produce robust, discriminative, and fusion-friendly representations which are very important in multimodal biometric authentication systems to enhance accuracy and reliability. Trust and confidence are crucial in multimodal biometric authentication systems utilizing deep learning, as the models operate as black boxes, handle irreversible biometric data, and make high-impact security decisions. This motivates the development of a secure, explainable, multimodal biometric authentication framework. The proposed system is a privacy-preserving and explainable multimodal biometric solution that combines deep learning, trust-adaptive fusion, and encrypted domain matching. It utilizes MobileNet for extracting discriminative features. A Trust Adaptive Fusion (TAF) Strategy adjusts the contribution of each modality based on its quality or confidence, enhancing the robustness against the noisy inputs. The fused features are secured using the Cheon-Kim-Kim-Song (CKKS) homomorphic encryption, without revealing the raw biometric data. Transparency is enhanced with the help of the Grad-CAM, which provides interpretability of the model’s decision. The proposed system is evaluated on the CASIA-FaceV5 and CASIA-FingerprintV5 datasets, demonstrates the low error rate of 0.0038 on fused feature representation.

State border monitoring represents a critical component of national security and migration management. The growth of globalization, international mobility, tourism, and cross-border trade has significantly increased the operational complexity of border control systems. Traditional monitoring methods based on manual document verification and visual inspection by officers are increasingly insufficient under conditions of high passenger traffic and growing security risks. As a result, modern border control infrastructures are progressively integrating advanced technological solutions aimed at improving identification accuracy, operational efficiency, and decision-making speed. Among these technologies, biometric identification systems play an increasingly important role. Biometric technologies allow individuals to be identified based on unique physiological or behavioral characteristics, reducing reliance on manual inspection and minimizing human error. In particular, facial recognition systems have become one of the most promising biometric approaches due to their non-contact nature, scalability, and ability to integrate seamlessly with existing video surveillance infrastructures. This paper analyzes the role of biometric identification technologies in modern border monitoring systems with particular emphasis on facial recognition methods. The study examines the technological evolution of border monitoring systems, discusses existing approaches to facial recognition, and evaluates the challenges associated with their implementation in real operational environments. Special attention is given to embedding-based recognition models and the impact of environmental factors such as illumination, pose variation, motion blur, and partial occlusion. The results highlight the importance of developing robust and adaptive recognition systems capable of maintaining reliable performance under dynamic border monitoring conditions.

Signature verification plays a crucial role in biometric authentication systems. Traditional local texture descriptors such as Local Binary Pattern (LBP), Local Directional Pattern (LDP), and Local Tetra Pattern (LTrP) exhibit limitations in capturing detailed gradient variations in signature strokes. This paper proposes a novel Local Gradient Hexa Pattern (LGHP) descriptor for robust feature extraction in offline signature verification. The proposed method encodes gradient magnitude and directional information into a six-pattern structure, enhancing discrimination capability. Experimental evaluation is performed on standard signature datasets using a Minimum Distance Classifier. Performance comparison demonstrates that the proposed LGHP achieves improved verification accuracy and reduced false acceptance rate compared to existing descriptors. The results confirm the effectiveness and robustness of the proposed approach for biometric authentication systems.

This paper proposes a novel radar-based framework for non-contact biometric identification through heart signal extraction, targeting secure and privacy-conscious identification scenarios. Traditional biometric methods, such as fingerprint and facial recognition, face challenges including privacy concerns, vulnerability to spoofing, and the requirement for close proximity or direct line-of-sight. Our framework addresses these issues by reconstructing electrocardiogram (ECG) signals from radar-extracted cardiac motion data and implementing an open-set person identification system. Specifically, the framework integrates ECGReconNet, a specialized deep learning model for reconstructing ECG signals from human chest wall displacement, the InceptionTime model enhanced with fixed-Class Anchor Clustering (fixed-CAC) loss for robust feature anchoring, and a hypersphere-based delineation method to differentiate known from unknown individuals. Experimental results on a public dataset demonstrate state-of-the-art performance, achieving 99.61% accuracy in closed-set identification (27 subjects) and 93.97% accuracy under challenging open-set conditions (14 known and 13 unknown subjects). However, the proposed approach exhibits limitations, including sensitivity to abrupt body movements and environmental noise, potential performance degradation under severe cardiac irregularities, and reduced efficacy with increased numbers of unknown identities.

IoT devices are increasingly targeted by cyber threats, where traditional passwords fail due to phishing and reuse, and raw biometric storage risk irreversible identity loss. This paper proposes an edge-centric biometric authentication system using lightweight deep learning (quantized MobileNetV2 CNN) for feature extraction from fingerprint or face data, cancelable transformations for revocable non-invertible templates, and AES-256 encryption for privacy. On-device processing on Raspberry Pi/ESP32 reduces latency and prevents data exposure. Evaluations on benchmark datasets yield >96% accuracy, EER <2%, and robust spoofing resistance, suitable for smart homes, healthcare, and IIoT

In the digital era, identity verification plays a crucial role in online services such as banking, healthcare, education, and government applications. Traditional identity management systems rely on centralized authorities that store sensitive user data, making them vulnerable to data breaches, identity theft, and unauthorized access. This paper proposes a Decentralized Identity Verification System (DID) based on blockchain technology that enables secure, transparent, and usercontrolled identity management. The proposed system utilizes blockchain networks and smart contracts to store encrypted identity records, ensuring immutability and privacy. Users maintain control over their identity credentials while institutions can verify authenticity without accessing sensitive personal data. The decentralized architecture eliminates dependency on centralized authorities and enhances trust, privacy, and security in digital identity verification.

Face-recognition technology is one of the most important advancements in the field of computer vision. They play a crucial role in many applications, including biometric authentication, surveillance, online security, and interactive web systems. The use of web-based solutions is increasing continuously. Therefore, accurate and fast recognition models employing few resources are required in real-world applications. However, because of the challenges related to such environments, including the lighting, occlusion, pose, and computing power of client devices, it is difficult to ascertain which model will be most successful in a real-life scenario. The purpose of this research is to compare four deep learning frameworks for face recognition, which are most widely used by scientists and software developers. FaceNet, SFace, OpenFace, and DeepFace have all been subjected to rigorous examinations to determine which one is the most suitable for work on the real-time web. As part of the assessment, a prototype application was created to enable the simulation of real-time applications. This solution enables both the upload of the test image and the group image to determine which person is the subject of the research. Subsequently, the model performance was tested under the conditions of pose, light, and occlusion variations. Performance was measured using the following features: accuracy, similarity distance, processing latency, and robustness. Therefore, the results show that there is no single best model compatible with all web-based applications, and the outcome fundamentally depends on the developer’s required accuracy and speed.

The iris recognition system is one of the most reliable biometric authentication systems owing to its individuality and permanence. With the introduction of deep learning concepts, especially CNNs, iris recognition models have improved tremendously in accuracy, robustness, and efficiency. We present an extensive survey summ2arizing the application of deep learning in iris recognition; covering datasets, feature extraction, architectures, and evaluation metrics. We conclude that while CNNs and transfer learning produce the best accuracy on well-constrained datasets, serious challenges lie with cross-sensor generalization and robustness under mobile or unconstrained environments. The new solutions that are generating attention, such as generation adversarial network-based augmentation and attention-driven architectures, are solution pathways to surmount data scarcity and further strengthen the adaptation capability. This survey is aimed at steering researchers and practitioners to various critical challenges and directions that have the most potential for future iris recognition systems.

Multi-modal biometric authentication systems face a critical trade-off: deep learning models achieve high accuracy but require substantial computational resources and lack interpretability making them unsuitable for edge devices and regulated environments. This paper addresses this gap by conducting the first comprehensive empirical comparison of Adaptive Neuro-Fuzzy Inference System (ANFIS) against five state-of-the-art deep learning architectures (CNN, DNN, RNN/LSTM, Transformer) for real-time multi-modal fusion of face, iris, and palmprint biometrics. Our novel contributions include: (1) a systematic evaluation framework demonstrating ANFIS achieves superior accuracy (98.1%) and lowest EER (1.5%) while requiring 93.7% less training time and 91.1% less inference time than Transformers; (2) statistical validation via 10-fold cross-validation proving ANFIS improvements are statistically significant (p<0.05); and (3) an adaptive weight adjustment mechanism enabling dynamic modality reweighting based on input quality. Results on three benchmark datasets (LFW, CASIA-Iris-V4, PolyU) with 20-user real-world validation demonstrate ANFIS as the optimal solution for resource-constrained, explainable biometric systems addressing critical deployment gaps in mobile, IoT, and privacy-sensitive applications.

Signing EEG-based biometric authentication system using multivariate Fourier-Bessel series expansion-based entropies

Speaker identification remains critical in biometric authentication systems, requiring robust feature extraction strategies that capture speaker-specific vocal characteristics. This study introduces a hybrid deep learning architecture integrating Convolutional Neural Networks (CNNs) with Reinforcement Learning (RL) for confidence-aware speaker identification. Two feature extraction methodologies were compared: Method 1 employed Mel-spectrogram representations (80 bins, 20-8000Hz) with self-attention mechanisms, while Method 2 utilized Continuous Wavelet Transform with Morlet wavelets (128 scales). Both methods were implemented as hybrid CNN-RL architectures and compared against CNN-only baselines. Frameworks were evaluated on LibriSpeech dev-clean dataset (2,703 audio files, 40 speakers) through stratified 5-fold cross-validation. ANOVA assessed discriminative capacity of 22 acoustic features. ANOVA revealed 21 of 22 features demonstrated significant discriminative power (p < 0.05), with entropy exhibiting the strongest effect (F = 39.79, η² = 0.37). Method 1 achieved 87.60% accuracy (95% CI: [83.60%, 91.95%]) and ROC-AUC of 99.54%. Method 2 attained 77.60% accuracy (95% CI: [73.12%, 82.08%]) with ROC-AUC of 98.21%. Ablation studies demonstrated that RL integration provided statistically significant improvements over CNN-only baselines (Method 1: +2.80%, p = 0.0142; Method 2: +3.40%, p = 0.0089) with reduced performance variability. Both architectures demonstrated balanced metrics with Matthews Correlation Coefficients exceeding 0.76. Q-learning integration enabled adaptive decision-making for classification uncertainty, with ablation studies confirming quantitative contributions over conventional CNNs. Results demonstrate that Mel-scale representations with attention mechanisms provide superior discriminative capacity compared to fixed-resolution wavelet transforms for robust biometric authentication.

The main aim of this project is to develop an IoT-based security system for residents which includes biometric authentication, plate recognition, and movement detection system. In this proposed method, the programming platforms such as Python and Arduino were used to develop and demonstrate the proposed system. The performance of the developed proposed system is evaluated by testing the system with several sample tests and from there, the performance was examined. The system performed well in recognizing the different persons and is capable of returning the correct output in almost all the face samples, as well as the plate number detection, which can successfully extract the string information from the pictures. It is observed that the system has an overall accuracy of 77% after considering several important factors that may affect the system’s performance.

The increasing number of intrusion and unauthorized room access incidents necessitates the development of a security system that is more intelligent, responsive, and resistant to forgery. Conventional security systems or card-based identification methods are highly vulnerable to theft, duplication, and misuse, highlighting the need for biometric authentication that can operate independently without reliance on centralized servers. In this study, an edge computing approach is implemented by utilizing a Raspberry Pi as a local processing unit capable of performing real-time data processing without requiring an internet connection. A RaspiCam camera is employed to capture facial images, which are then detected using the YOLO (You Only Look Once) algorithm to achieve fast and efficient face detection. The detected faces are subsequently recognized using a Convolutional Neural Network (CNN). To further enhance security, a voice-based authentication stage is applied using the Mel-Frequency Cepstral Coefficients (MFCC) method to verify user identity. Experimental results demonstrate that the face recognition module achieves an accuracy of up to 96.5% under normal lighting conditions, while the voice recognition module reaches an accuracy of 94.7% in a quiet environment with a signal-to-noise ratio of 30 dB. Both authentication processes are executed sequentially with an average processing time of less than 500 ms, enabling real-time operation on an edge device. This approach provides a robust two-factor biometric authentication system that can be deployed without dependence on cloud-based services.