Sk Mahmudul Hassan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sk Mahmudul Hassan has a total of 52 co-authors. Here are the top co-authors by number of shared publications:
- Arnab Kumar Maji
Department of Information Technology, School of Technology, NEHU, Shillong, India
6 shared publications Kumar Sekhar Roy
Affiliation Manipal Institute of Technology Bengaluru, Manipal Academy of Higher Education, Manipal, Karnataka, India
4 shared publications- Kumar Sekhar Roy
Manipal Institute of Technology Bengaluru, Manipal Academy of Higher Education, Manipal, India
4 shared publications - Ruhul Amin Hazarika
Department of Computer Engineering, Modeling, Electronics, and Systems (DIMES), University of Calabria, Rende, Cosenza, Italy and Department of Computer Science and Engineering, Indian Institute of Information Technology Design and Manufacturing, Kurnool, Andhra Pradesh, India
4 shared publications Arnab Kumar Maji
North Eastern Hill University
3 shared publications- Elżbieta Jasińska
Department of Integrated Geodesy and Cartography, Faculty of Geo-Data Science, Geodesy and Environmental Engineering, AGH University of Krakow, al. Mickiewicza 30, 30-059 Kraków, Poland
3 shared publications
Latest Publications
- Open Access
- 27 October 2026
- Published by Springer Nature in Discover Applied Sciences
Lightweight transformer based WT-MobileViT framework for plant disease detection
The rapid spread of plant diseases significantly threatens global food security. Early and accurate detection of these diseases is crucial for effective disease management and crop protection. Manual identification of diseases requires expert knowledge in the field, and finding people with domain knowledge is challenging. In recent years, computer vision-based machine learning techniques have been proposed by researchers to overcome the challenge. In this paper, we have proposed a WT-MobileViT, a lightweight hybrid model that integrates frequency domain information with the MobileViT framework for the classification of plant diseases. The proposed method first performs a one-level Haar Discrete Wavelet Transform (DWT) to decompose the input image into low-frequency and high-frequency sub-bands, preserving complementary structural and texture information, and concatenates with the original RGB image to construct a multi-frequency feature representation. In this article, we have created a tea leaf disease dataset with RGB images in field conditions. The proposed model was experimented on four datasets and achieved an accuracy rate of 99.44% in tea leaf images, 99.68% in the rice dataset, 99.17% in the bean dataset, and 99.21% in plantvillage dataset. The experimental results show that the proposed model outperforms several state-of-the-art deep learning models with fewer parameters.
- 27 October 2026
- Published by Springer Nature in Discover Applied Sciences
Open AccessLightweight transformer based WT-MobileViT framework for plant disease detection
The rapid spread of plant diseases significantly threatens global food security. Early and accurate detection of these diseases is crucial for effective disease management and crop protection. Manual identification of diseases requires expert knowledge in the field, and finding people with domain knowledge is challenging. In recent years, computer vision-based machine learning techniques have been proposed by researchers to overcome the challenge. In this paper, we have proposed a WT-MobileViT, a lightweight hybrid model that integrates frequency domain information with the MobileViT framework for the classification of plant diseases. The proposed method first performs a one-level Haar Discrete Wavelet Transform (DWT) to decompose the input image into low-frequency and high-frequency sub-bands, preserving complementary structural and texture information, and concatenates with the original RGB image to construct a multi-frequency feature representation. In this article, we have created a tea leaf disease dataset with RGB images in field conditions. The proposed model was experimented on four datasets and achieved an accuracy rate of 99.44% in tea leaf images, 99.68% in the rice dataset, 99.17% in the bean dataset, and 99.21% in plantvillage dataset. The experimental results show that the proposed model outperforms several state-of-the-art deep learning models with fewer parameters.
- Open Access
- 12 July 2026
- Published by Springer Nature
- p. 33-44
AI-Powered Food Supply Chain: Data-Driven Decision Making for Fraud Detection
The complexity besides scalability of the world’s food supply chain has both increased in recent years. Environmental, demographic, and economic shifts have a major impact on the efficiency of food production systems from farm to table. Human health could be put at risk as a result of these developments, which could lead to an upsurge in food fraud besides safety hazards. To lessen the impact of these risks, the food supply chain might use AI technologies. A lot of what AI systems perform is mysterious since it’s hard to describe, even if its usage has been on the rise in many sectors like precision nutrition, self-driving vehicles, precision safety. For the purpose of detecting food fraud, the study presented a new attention-integrating mechanism called CNN-LSTM-SE, which is based on the CNN-Long Short-Term Memory-squeeze-and-excitation network model. To improve classification accuracy, Binary Manta Ray Foraging (BMRF) ideally selects the LSTM parameters. Data used in the case study came from a food fraud dataset that included reports of adulteration and fraud pulled from two sources: the Rapid Alert Scheme for Food and Feed and the Frugally Motivated Contamination database.
- 12 July 2026
- Published by Springer Nature
- p. 33-44
Open AccessAI-Powered Food Supply Chain: Data-Driven Decision Making for Fraud Detection
The complexity besides scalability of the world’s food supply chain has both increased in recent years. Environmental, demographic, and economic shifts have a major impact on the efficiency of food production systems from farm to table. Human health could be put at risk as a result of these developments, which could lead to an upsurge in food fraud besides safety hazards. To lessen the impact of these risks, the food supply chain might use AI technologies. A lot of what AI systems perform is mysterious since it’s hard to describe, even if its usage has been on the rise in many sectors like precision nutrition, self-driving vehicles, precision safety. For the purpose of detecting food fraud, the study presented a new attention-integrating mechanism called CNN-LSTM-SE, which is based on the CNN-Long Short-Term Memory-squeeze-and-excitation network model. To improve classification accuracy, Binary Manta Ray Foraging (BMRF) ideally selects the LSTM parameters. Data used in the case study came from a food fraud dataset that included reports of adulteration and fraud pulled from two sources: the Rapid Alert Scheme for Food and Feed and the Frugally Motivated Contamination database.
- Open Access
- 15 May 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (5) , e0348293
Design and evaluation of a resilient IBN architecture: Integrating post-quantum cryptography with adaptive threat detection using machine learning
As the domain of network security keeps on evolving rapidly, especially in sensitive areas such as healthcare systems, the demand for reliable device verification, controlling access, and spotting threats is growing sharply. This paper presents the design, implementation, and systematic evaluation of an improved Intent-Based Networking (IBN) system that integrates post-quantum cryptography, certificate-based identity management, and machine learning-based anomaly detection within a unified framework. The system incorporates SPHINCS+ post-quantum digital signatures for quantum-resilient authentication, X.509 certificate lifecycle management for establishing device trust, and hardware-aware cryptographic operations to maintain efficiency. It further enforces fine-grained access policies using Role-Based Access Control (RBAC) augmented with Multi-Factor Authentication (MFA), ensuring strong access governance across network segments. For early threat detection, machine learning models such as Isolation Forest and MiniBatch KMeans are employed to learn communication patterns and detect anomalous device behavior. Additionally, event logs are maintained using asynchronous, hash-chained logging mechanisms inspired by blockchain principles, ensuring auditability and data integrity. To address evaluation transparency and rigor, the framework is assessed using a controlled prototype testbed with explicitly defined traffic features and reproducible experimental settings. The evaluation considers cryptographic correctness, access control performance, anomaly detection capability, and scalability under increasing workloads. Experimental results demonstrate 100% success in post-quantum signature generation and verification, effective anomaly detection with no observed false negatives in the evaluated scenarios, and stable log-processing throughput as the number of events grows. Importantly, this work does not claim novelty in individual components, but contributes through the system-level integration and empirical evaluation of a quantum-safe, ML-assisted IBN security architecture. The findings highlight key trade-offs between security enforcement and usability, while also identifying limitations such as certificate expiry handling gaps, conservative policy behavior, and lack of large-scale statistical validation. These observations establish a reproducible baseline and motivate future work toward statistically rigorous validation, real-world deployment, and adaptive policy optimization.
- 15 May 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (5) , e0348293
Open AccessDesign and evaluation of a resilient IBN architecture: Integrating post-quantum cryptography with adaptive threat detection using machine learning
As the domain of network security keeps on evolving rapidly, especially in sensitive areas such as healthcare systems, the demand for reliable device verification, controlling access, and spotting threats is growing sharply. This paper presents the design, implementation, and systematic evaluation of an improved Intent-Based Networking (IBN) system that integrates post-quantum cryptography, certificate-based identity management, and machine learning-based anomaly detection within a unified framework. The system incorporates SPHINCS+ post-quantum digital signatures for quantum-resilient authentication, X.509 certificate lifecycle management for establishing device trust, and hardware-aware cryptographic operations to maintain efficiency. It further enforces fine-grained access policies using Role-Based Access Control (RBAC) augmented with Multi-Factor Authentication (MFA), ensuring strong access governance across network segments. For early threat detection, machine learning models such as Isolation Forest and MiniBatch KMeans are employed to learn communication patterns and detect anomalous device behavior. Additionally, event logs are maintained using asynchronous, hash-chained logging mechanisms inspired by blockchain principles, ensuring auditability and data integrity. To address evaluation transparency and rigor, the framework is assessed using a controlled prototype testbed with explicitly defined traffic features and reproducible experimental settings. The evaluation considers cryptographic correctness, access control performance, anomaly detection capability, and scalability under increasing workloads. Experimental results demonstrate 100% success in post-quantum signature generation and verification, effective anomaly detection with no observed false negatives in the evaluated scenarios, and stable log-processing throughput as the number of events grows. Importantly, this work does not claim novelty in individual components, but contributes through the system-level integration and empirical evaluation of a quantum-safe, ML-assisted IBN security architecture. The findings highlight key trade-offs between security enforcement and usability, while also identifying limitations such as certificate expiry handling gaps, conservative policy behavior, and lack of large-scale statistical validation. These observations establish a reproducible baseline and motivate future work toward statistically rigorous validation, real-world deployment, and adaptive policy optimization.
- Open Access
- 13 May 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Dynamic community evolution analysis for performance optimization in large scale RFID networks
Handling Large scale RFID network is most important for Tracking and Locating objects in the Intelligent Transportation System and IoT based RFID networks. This research work focuses to build the road map with RFID support without interference of satellite. The cluster-based RFID network can be supported to enlarge the RFID network. In this paper, proposes the Novel Community Evolution Analysis (NCEA) method for selecting cluster heads, designed to tackle both strong and weak events within the network. The traditional clustering method may miss some weak events occurrences and its leads to impact the efficiency of cluster head selection of the cluster network. To identify this problem, our method incorporates the required strong and weak events. This proposed method concentrates the energy management and cluster head selection management accurately. Also, the cluster head selection may happen through high energy node and close neighbor node to all the participated nodes in the cluster. The cluster head selection can be happened by different events such as form, disappear, shrink, expand, split, merge. The NCEA approach demonstrates an accuracy of 98%, a vulnerability rate of 20%, a success rate of 89%, a latency of 11.4 s, and a throughput of 93%. These results highlight the effectiveness of our method in enhancing the energy efficiency and overall performance of large-scale RFID networks.
- 13 May 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessDynamic community evolution analysis for performance optimization in large scale RFID networks
Handling Large scale RFID network is most important for Tracking and Locating objects in the Intelligent Transportation System and IoT based RFID networks. This research work focuses to build the road map with RFID support without interference of satellite. The cluster-based RFID network can be supported to enlarge the RFID network. In this paper, proposes the Novel Community Evolution Analysis (NCEA) method for selecting cluster heads, designed to tackle both strong and weak events within the network. The traditional clustering method may miss some weak events occurrences and its leads to impact the efficiency of cluster head selection of the cluster network. To identify this problem, our method incorporates the required strong and weak events. This proposed method concentrates the energy management and cluster head selection management accurately. Also, the cluster head selection may happen through high energy node and close neighbor node to all the participated nodes in the cluster. The cluster head selection can be happened by different events such as form, disappear, shrink, expand, split, merge. The NCEA approach demonstrates an accuracy of 98%, a vulnerability rate of 20%, a success rate of 89%, a latency of 11.4 s, and a throughput of 93%. These results highlight the effectiveness of our method in enhancing the energy efficiency and overall performance of large-scale RFID networks.
- Open Access
- 18 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16
Epistemic uncertain computing for intrusion detection with explainability & multi-criteria optimization using AA-NLS and GRUSO-GRU
Owing to incomplete and inaccurate data, intrusion detection in networks is uncertain. The out-of-order packet arrival uncertainty was not analyzed in the existing works, thereby causing false positives in intrusion detection. So, in this paper, a novel Algebraic Average-based Neutrosophic Logic System (AA-NLS)-based out-of-order uncertainty analysis is proposed. Firstly, the nodes are initialized, and the data is sensed. Then, for intrusion detection, the Network Intrusion Detection System (NIDS) is trained. Here, the data is collected and pre-processed. At the time of pre-processing, Weighted Distance Error Function-based K-Nearest Neighbour (WDEFKNN) is utilized to impute the uncertainty caused by missing values. Further, utilizing the Density-Based Beta Distribution Soergel Spatial Clustering of Applications with Noise (DBBDSSCAN), the behavioural patterns are grouped. Then, the features are extracted, and by utilizing AA-NLS, the epistemic uncertainty caused by out-of-order packet arrival is analyzed . In the meantime, by utilizing the Hidden Laplace Witten Bell Markov Model (HLWBMM), the temporal patterns are assessed. Then, the intrusion is detected by the Generalized Riccati Uniform Scaling Orthogonal-based Gated Recurrent Unit (GRUSO-GRU). Here, utilizing the Pareto Entropy-based SHapley Additive exPlanation (PESHAP) model, the explainability of the intrusion detection is carried out. If there is no intrusion, then data transmission is continued. And, if there is intrusion, then the gathered alerts are prioritized and transmitted utilizing the multi-criteria-based Log-Cosh Fennec Fox Optimization Algorithm (LCFFOA). Hence, the intrusion is detected by the proposed framework with an accuracy of 99.299% under the full multi-class setting, demonstrating effective performance compared to prevailing Deep Learning (DL)-based NIDS approaches.
- 18 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16
Open AccessEpistemic uncertain computing for intrusion detection with explainability & multi-criteria optimization using AA-NLS and GRUSO-GRU
Owing to incomplete and inaccurate data, intrusion detection in networks is uncertain. The out-of-order packet arrival uncertainty was not analyzed in the existing works, thereby causing false positives in intrusion detection. So, in this paper, a novel Algebraic Average-based Neutrosophic Logic System (AA-NLS)-based out-of-order uncertainty analysis is proposed. Firstly, the nodes are initialized, and the data is sensed. Then, for intrusion detection, the Network Intrusion Detection System (NIDS) is trained. Here, the data is collected and pre-processed. At the time of pre-processing, Weighted Distance Error Function-based K-Nearest Neighbour (WDEFKNN) is utilized to impute the uncertainty caused by missing values. Further, utilizing the Density-Based Beta Distribution Soergel Spatial Clustering of Applications with Noise (DBBDSSCAN), the behavioural patterns are grouped. Then, the features are extracted, and by utilizing AA-NLS, the epistemic uncertainty caused by out-of-order packet arrival is analyzed . In the meantime, by utilizing the Hidden Laplace Witten Bell Markov Model (HLWBMM), the temporal patterns are assessed. Then, the intrusion is detected by the Generalized Riccati Uniform Scaling Orthogonal-based Gated Recurrent Unit (GRUSO-GRU). Here, utilizing the Pareto Entropy-based SHapley Additive exPlanation (PESHAP) model, the explainability of the intrusion detection is carried out. If there is no intrusion, then data transmission is continued. And, if there is intrusion, then the gathered alerts are prioritized and transmitted utilizing the multi-criteria-based Log-Cosh Fennec Fox Optimization Algorithm (LCFFOA). Hence, the intrusion is detected by the proposed framework with an accuracy of 99.299% under the full multi-class setting, demonstrating effective performance compared to prevailing Deep Learning (DL)-based NIDS approaches.
- Open Access
- 8 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Federated learning with continual update for privacy-preserving clinical event prediction across distributed hospitals using MCN-GNN
Federated Learning (FL) enables accurate and secure Clinical Event Prediction (CEP) across distributed hospitals. However, the prevailing works overlooked the catastrophic forgetting during the global update. Therefore, a Meta Experience Polynomial Decay-based Replay (MEPDR)-centric continual update is proposed. Initially, the hospitals (local model) register and log into the blockchain. Then, to train the CEP model, data collection, pre-processing, and feature extraction are performed. Further, the Temporal-Causal Graph (TCG) is constructed. Afterward, the node matrix is created, and the CEP is done using Mean-Centering Normalization-based Graph Neural Network (MCN-GNN). The model’s gradients are further preserved using the Homomorphic Robust Log Scaling-based Encryption (HRLSE). Next, the hospitals are authenticated using the Exponential Probing Digital Signature Algorithm (ExPrDSA). Thereafter, in the global model, the aggregation is performed using the Calinski–Harabasz Index with Zhonghua Distance-based K-Means Clustering (CHIZD-KMC), followed by global CEP. After that, during the global update, the MEPDR-based continual learning is carried out in each local model. Also, the transactions are stored in the blockchain to enhance traceability. Thus, the proposed system effectively predicted the clinical events with an accuracy of 98.97%, outperforming existing works.
- 8 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessFederated learning with continual update for privacy-preserving clinical event prediction across distributed hospitals using MCN-GNN
Federated Learning (FL) enables accurate and secure Clinical Event Prediction (CEP) across distributed hospitals. However, the prevailing works overlooked the catastrophic forgetting during the global update. Therefore, a Meta Experience Polynomial Decay-based Replay (MEPDR)-centric continual update is proposed. Initially, the hospitals (local model) register and log into the blockchain. Then, to train the CEP model, data collection, pre-processing, and feature extraction are performed. Further, the Temporal-Causal Graph (TCG) is constructed. Afterward, the node matrix is created, and the CEP is done using Mean-Centering Normalization-based Graph Neural Network (MCN-GNN). The model’s gradients are further preserved using the Homomorphic Robust Log Scaling-based Encryption (HRLSE). Next, the hospitals are authenticated using the Exponential Probing Digital Signature Algorithm (ExPrDSA). Thereafter, in the global model, the aggregation is performed using the Calinski–Harabasz Index with Zhonghua Distance-based K-Means Clustering (CHIZD-KMC), followed by global CEP. After that, during the global update, the MEPDR-based continual learning is carried out in each local model. Also, the transactions are stored in the blockchain to enhance traceability. Thus, the proposed system effectively predicted the clinical events with an accuracy of 98.97%, outperforming existing works.
- Open Access
- 1 January 2026
- Published by Institution of Engineering and Technology (IET) in IET Quantum Communication
- Vol. 7 (1) , e70041
Adaptive Machine Learning‐Enhanced Teleportation‐Based Quantum Key Distribution
Hybrid quantum networks will be the communication backbone of the future quantum internet, linking together heterogeneous discrete‐variable (DV) and continuous‐variable (CV) quantum systems. However, current quantum key distribution protocols assume a static basis and are not designed for dynamically changing hybrid quantum environments. This paper suggests an adaptive machine learning‐assisted teleportation quantum key distribution framework that integrates the following steps into a quantum‐classical architecture: hybrid DV–CV communication, quantum teleportation, adaptive basis selection and neural‐assisted quantum error correction. Optimal measurement basis prediction, by means of a supervised multilayer perceptron (MLP), using six real‐time channel descriptors, allows for intelligent protocol adaptation under heterogeneous channel conditions. The framework proposed was realized in Qiskit Aer and Scikit‐learn and tested with 6000 different realisations of the different channels. The classification accuracy was 99.44% while the average inference latency was only 0.0011 ms per channel sample, making it possible to use the adaptive basis selector in real time. The teleportation fidelity of AMT‐QKD was 0.899 at a representative channel noise level of 10%, with a quantum bit error rate (QBER) of 0.088, a secret key rate of 0.404 and an eavesdropper mutual information of only . They are shown to be more robust, have less information leakage, and be more secure in generating keys under realistic channel impairments when compared to other protocols, including BB84, Enhanced BB84, B92, BBM92, E91 and CV‐QKD. The findings provide a robust framework for the deployment of AMT‐QKD in secure communication in future hybrid quantum networks and quantum internet applications.
- 1 January 2026
- Published by Institution of Engineering and Technology (IET) in IET Quantum Communication
- Vol. 7 (1) , e70041
Open AccessAdaptive Machine Learning‐Enhanced Teleportation‐Based Quantum Key Distribution
Hybrid quantum networks will be the communication backbone of the future quantum internet, linking together heterogeneous discrete‐variable (DV) and continuous‐variable (CV) quantum systems. However, current quantum key distribution protocols assume a static basis and are not designed for dynamically changing hybrid quantum environments. This paper suggests an adaptive machine learning‐assisted teleportation quantum key distribution framework that integrates the following steps into a quantum‐classical architecture: hybrid DV–CV communication, quantum teleportation, adaptive basis selection and neural‐assisted quantum error correction. Optimal measurement basis prediction, by means of a supervised multilayer perceptron (MLP), using six real‐time channel descriptors, allows for intelligent protocol adaptation under heterogeneous channel conditions. The framework proposed was realized in Qiskit Aer and Scikit‐learn and tested with 6000 different realisations of the different channels. The classification accuracy was 99.44% while the average inference latency was only 0.0011 ms per channel sample, making it possible to use the adaptive basis selector in real time. The teleportation fidelity of AMT‐QKD was 0.899 at a representative channel noise level of 10%, with a quantum bit error rate (QBER) of 0.088, a secret key rate of 0.404 and an eavesdropper mutual information of only . They are shown to be more robust, have less information leakage, and be more secure in generating keys under realistic channel impairments when compared to other protocols, including BB84, Enhanced BB84, B92, BBM92, E91 and CV‐QKD. The findings provide a robust framework for the deployment of AMT‐QKD in secure communication in future hybrid quantum networks and quantum internet applications.
- Open Access
- 3 November 2025
- Published by Springer Nature in Earth Systems and Environment
- Vol. 10 (5) , 5383-5399
AquaFusionNet: A Deep Feature Fusion and Attention-Based Approach for Accurate Underwater Defect Detection
Underwater infrastructures such as pipelines, ship hulls, and offshore platforms are critical to marine operations but are highly vulnerable to biofouling, structural corrosion, and vegetation overgrowth, leading to increased maintenance costs and environmental hazards. However, visual inspection of these structures remains challenging due to low visibility, uneven lighting, and complex textured surfaces that limit the effectiveness of both traditional and purely deep learning-based approaches. In this work, we introduce AquaFusionNet, a hybrid defect classification framework that seamlessly integrates embeddings from four state-of-the-art pre-trained backbones (EfficientNet-B0, ResNet-50, SENet-50, and Vision Transformer) with complementary traditional descriptors including colour histograms, histogram of oriented gradients (HOG), local binary patterns (LBP), edge density, and gradient statistics via a trainable attention module. This attention mechanism dynamically weights each feature channel, allowing the model to emphasise the most informative cues while preserving fine-scale details under variable turbidity and illumination. We evaluate AquaFusionNet on a curated dataset of 2,228 underwater images spanning three defect categories, with 445 images held out for testing. Our model achieves 98.43% accuracy, 98.18% precision, 97.79% recall, a 96.06% intersection-over-union, and a 97.97% F $$_1$$ -score, outperforming eleven strong baselines, including ResNet-152 and EfficientNet-B0, by a substantial margin. These results demonstrate AquaFusionNet’s robustness and generalisability, paving the way for real-time, automated underwater inspection systems that can significantly enhance operational safety and reduce maintenance costs across marine industries. Graphic This graphical abstract provides a clear overview of the AquaFusionNet research, illustrating a streamlined workflow moving from data acquisition through final results and conclusions. The dataset comprises 2,228 underwater images categorized into biofouling, holes, and vegetation, sourced from the ICIP MVI-2024 challenge and split into training, validation, and testing subsets with a ratio of 60:20:20. The analysis phase emphasizes preprocessing steps combined with traditional descriptors, including color histograms, texture patterns, and gradient-based features, complementing the extracted deep features. The AquaFusionNet model uniquely integrates embeddings from four advanced, lightweight backbones, EfficientNet-B0, ResNet-50, SE-ResNet-50, and Vision Transformer (ViT-B/16) with traditional features, using a dynamic trainable attention mechanism. This attention module adaptively emphasizes the most informative features, which are then classified via a two-layer multilayer perceptron (MLP). Evaluation of the model reveals exceptional performance metrics, achieving an accuracy of 98.43%, an F1-score of 97.97%, and an intersection-over-union (IoU) of 96.06%, surpassing eleven state-of-the-art models on a test set of 445 images. The final conclusion highlights AquaFusionNet as an efficient, deployment-ready framework capable of robustly detecting underwater defects, significantly enhancing the safety and reducing the costs associated with marine inspections.- 3 November 2025
- Published by Springer Nature in Earth Systems and Environment
- Vol. 10 (5) , 5383-5399
Open AccessAquaFusionNet: A Deep Feature Fusion and Attention-Based Approach for Accurate Underwater Defect Detection
Underwater infrastructures such as pipelines, ship hulls, and offshore platforms are critical to marine operations but are highly vulnerable to biofouling, structural corrosion, and vegetation overgrowth, leading to increased maintenance costs and environmental hazards. However, visual inspection of these structures remains challenging due to low visibility, uneven lighting, and complex textured surfaces that limit the effectiveness of both traditional and purely deep learning-based approaches. In this work, we introduce AquaFusionNet, a hybrid defect classification framework that seamlessly integrates embeddings from four state-of-the-art pre-trained backbones (EfficientNet-B0, ResNet-50, SENet-50, and Vision Transformer) with complementary traditional descriptors including colour histograms, histogram of oriented gradients (HOG), local binary patterns (LBP), edge density, and gradient statistics via a trainable attention module. This attention mechanism dynamically weights each feature channel, allowing the model to emphasise the most informative cues while preserving fine-scale details under variable turbidity and illumination. We evaluate AquaFusionNet on a curated dataset of 2,228 underwater images spanning three defect categories, with 445 images held out for testing. Our model achieves 98.43% accuracy, 98.18% precision, 97.79% recall, a 96.06% intersection-over-union, and a 97.97% F $$_1$$ -score, outperforming eleven strong baselines, including ResNet-152 and EfficientNet-B0, by a substantial margin. These results demonstrate AquaFusionNet’s robustness and generalisability, paving the way for real-time, automated underwater inspection systems that can significantly enhance operational safety and reduce maintenance costs across marine industries. Graphic This graphical abstract provides a clear overview of the AquaFusionNet research, illustrating a streamlined workflow moving from data acquisition through final results and conclusions. The dataset comprises 2,228 underwater images categorized into biofouling, holes, and vegetation, sourced from the ICIP MVI-2024 challenge and split into training, validation, and testing subsets with a ratio of 60:20:20. The analysis phase emphasizes preprocessing steps combined with traditional descriptors, including color histograms, texture patterns, and gradient-based features, complementing the extracted deep features. The AquaFusionNet model uniquely integrates embeddings from four advanced, lightweight backbones, EfficientNet-B0, ResNet-50, SE-ResNet-50, and Vision Transformer (ViT-B/16) with traditional features, using a dynamic trainable attention mechanism. This attention module adaptively emphasizes the most informative features, which are then classified via a two-layer multilayer perceptron (MLP). Evaluation of the model reveals exceptional performance metrics, achieving an accuracy of 98.43%, an F1-score of 97.97%, and an intersection-over-union (IoU) of 96.06%, surpassing eleven state-of-the-art models on a test set of 445 images. The final conclusion highlights AquaFusionNet as an efficient, deployment-ready framework capable of robustly detecting underwater defects, significantly enhancing the safety and reducing the costs associated with marine inspections. - Open Access
- 23 August 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-17
Multi-kernel inception-enhanced vision transformer for plant leaf disease recognition
The timely and precise identification of diseases in plants is essential for efficient disease control and safeguarding of crops. Manual identification of diseases requires expert knowledge in the field, and finding people with domain knowledge is challenging. To overcome the challenge, computer vision-based machine learning techniques have been proposed by the researchers in recent years. Most of these solutions with the standard convolutional neural network (CNN) approaches use uniform background laboratory setup leaf images to identify the diseases. However, only a few works considered real-field images in their work. Therefore, there is a need for a robust CNN architecture that can identify the diseases in plants in both laboratory and real-field conditioned images. In this paper, we have proposed an Inception-Enhanced Vision Transformer (IEViT) architecture to identify diseases in plants. The proposed IEViT architecture extracts local as well as global features, which improves feature learning. The use of multiple filters with different kernel sizes efficiently uses computing resources to extract relevant features without the need for deeper networks. The robustness of the proposed architecture is established by hyper-parameter tuning and comparison with state-of-the-art. In the experiment, we consider five datasets with both laboratory-conditioned and real-field conditioned images. From the experimental results, we see that the proposed model outperforms state-of-the-art deep learning models with fewer parameters. The proposed model achieves an accuracy rate of 99.23% for the apple leaf dataset, 99.70% for the rice dataset, 97.02% for the ibean dataset, 76.51% for the cassava leaf dataset, and 99.41% for the plantvillage dataset.
- 23 August 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-17
Open AccessMulti-kernel inception-enhanced vision transformer for plant leaf disease recognition
The timely and precise identification of diseases in plants is essential for efficient disease control and safeguarding of crops. Manual identification of diseases requires expert knowledge in the field, and finding people with domain knowledge is challenging. To overcome the challenge, computer vision-based machine learning techniques have been proposed by the researchers in recent years. Most of these solutions with the standard convolutional neural network (CNN) approaches use uniform background laboratory setup leaf images to identify the diseases. However, only a few works considered real-field images in their work. Therefore, there is a need for a robust CNN architecture that can identify the diseases in plants in both laboratory and real-field conditioned images. In this paper, we have proposed an Inception-Enhanced Vision Transformer (IEViT) architecture to identify diseases in plants. The proposed IEViT architecture extracts local as well as global features, which improves feature learning. The use of multiple filters with different kernel sizes efficiently uses computing resources to extract relevant features without the need for deeper networks. The robustness of the proposed architecture is established by hyper-parameter tuning and comparison with state-of-the-art. In the experiment, we consider five datasets with both laboratory-conditioned and real-field conditioned images. From the experimental results, we see that the proposed model outperforms state-of-the-art deep learning models with fewer parameters. The proposed model achieves an accuracy rate of 99.23% for the apple leaf dataset, 99.70% for the rice dataset, 97.02% for the ibean dataset, 76.51% for the cassava leaf dataset, and 99.41% for the plantvillage dataset.
- Open Access
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71 (4) , 10152-10159
QSMIT: A Quantum Secure Medical Image Transmission Using Sphincs+ With DICOM
As quantum computing capabilities continue to evolve, the need for quantum-secure communication protocols is becoming increasingly essential, and ongoing research in this area is crucial to maintain the security and privacy of digital information in healthcare and beyond. Considering the integrity of such data, we proposed a secure medical image transmission using a post-quantum digital signature scheme known as Sphincs+, as suggested by the National Institute of Standards and Technology. Sphincs+ offers strong security against quantum attacks, ensuring long-term data protection. It is stateless, avoids issues with key reuse, and provides efficient, scalable digital signatures. Its robustness and flexibility make it ideal for diverse applications, maintaining integrity even as quantum computing advances. We transmit a digitally signed image using DICOM encoding and Quantum Fourier Transform for transmission. DICOM ensures standardized, interoperable management of medical images, integrating devices and systems for accurate diagnosis for modern consumer healthcare workflows and record-keeping. The Quantum Fourier Transform offers exponential speedup and higher precision over the classical Fast Fourier Transform, particularly benefiting applications in signal processing, data encryption, and environmental sensing. Advanced mistake detection and repair techniques, as well as extraordinarily fast data transmission, are possible with quantum communication. Additionally, it preserves the confidentiality and integrity of data while lowering the possibility of data loss or corruption during transmission. Some popular performance measuring parameters, such as PNR, MSE, SSIM, etc., are used to validate the proposed framework and observe a convincing result.
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71 (4) , 10152-10159
Open AccessQSMIT: A Quantum Secure Medical Image Transmission Using Sphincs+ With DICOM
As quantum computing capabilities continue to evolve, the need for quantum-secure communication protocols is becoming increasingly essential, and ongoing research in this area is crucial to maintain the security and privacy of digital information in healthcare and beyond. Considering the integrity of such data, we proposed a secure medical image transmission using a post-quantum digital signature scheme known as Sphincs+, as suggested by the National Institute of Standards and Technology. Sphincs+ offers strong security against quantum attacks, ensuring long-term data protection. It is stateless, avoids issues with key reuse, and provides efficient, scalable digital signatures. Its robustness and flexibility make it ideal for diverse applications, maintaining integrity even as quantum computing advances. We transmit a digitally signed image using DICOM encoding and Quantum Fourier Transform for transmission. DICOM ensures standardized, interoperable management of medical images, integrating devices and systems for accurate diagnosis for modern consumer healthcare workflows and record-keeping. The Quantum Fourier Transform offers exponential speedup and higher precision over the classical Fast Fourier Transform, particularly benefiting applications in signal processing, data encryption, and environmental sensing. Advanced mistake detection and repair techniques, as well as extraordinarily fast data transmission, are possible with quantum communication. Additionally, it preserves the confidentiality and integrity of data while lowering the possibility of data loss or corruption during transmission. Some popular performance measuring parameters, such as PNR, MSE, SSIM, etc., are used to validate the proposed framework and observe a convincing result.