Munindra Lunagaria
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Munindra Lunagaria has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
- Deepak Kumar Verma
Marwadi University
4 shared publications - Krunal Vaghela
Department of Computer Engineering, Marwadi University, Rajkot, Gujarat, India
4 shared publications - Pratikkumar Chauhan
Department of Computer Engineering, Marwadi University, Rajkot, Gujarat, India
4 shared publications - Ghanshyam G. Tejani
Department of Research Analytics, Saveetha Dental College and Hospitals, Saveetha Institute of Medical and Technical Sciences, Saveetha University, Chennai, 600077 India
3 shared publications - Ahmad Raza Khan
Department of Information Technology, College of Computer and Information Sciences, Majmaah University, Majmaah, Saudi Arabia
2 shared publications - Sunil Kumar Sharma
Majmaah University
2 shared publications
Latest Publications
- Open Access
- 6 January 2026
- Published by Elsevier in Alexandria Engineering Journal
- Vol. 135, 36-63
Enhancing biometric authentication privacy and security: A synergistic approach using cancelable biometrics and federated learning
- 6 January 2026
- Published by Elsevier in Alexandria Engineering Journal
- Vol. 135, 36-63
Open AccessEnhancing biometric authentication privacy and security: A synergistic approach using cancelable biometrics and federated learning
- 25 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Standards Magazine
- Vol. PP (24712825) , 1-13
Optimizing CNN-Based Video Summarization With Digital Twin Framework for Adaptive Probability Thresholding
The rapid growth of video data necessitates efficient summarization techniques to extract critical insights while minimizing computational overhead. This study investigates three pre-trained CNN models—AlexNet (61 M parameters), GoogLeNet (6.8M), and SqueezeNet (1.24M)—with respective inference speeds of 23.8 ms/frame, 18.4 ms/frame, and 12.6 ms/frame. These computational differences significantly affect real-time summarization performance. A probability-driven thresholding mechanism was introduced to optimize object detection, effectively reducing false positives and enhancing classification accuracy. Among the models, GoogLeNet achieved the best trade-off between accuracy (98.96%) and computational efficiency, while AlexNet attained the highest detection accuracy (99.12%) but exhibited a higher false detection rate (5.6%). SqueezeNet, though lightweight and resource-efficient, reached an accuracy of 94.26% but faced limitations in fine-grained recognition tasks. Probability versus correct detection analysis indicated that a threshold of 0.7 maximized accuracy while reducing misclassifications. The proposed adaptive thresholding framework demonstrates strong applicability in real-world scenarios such as wildlife monitoring, intelligent surveillance, and IoT-enabled video analytics. By bridging CNN-based architectures with scalable IoT-driven solutions, this research advances the development of robust, accurate, and efficient video summarization systems for next-generation intelligent environments.
- 25 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Standards Magazine
- Vol. PP (24712825) , 1-13
Optimizing CNN-Based Video Summarization With Digital Twin Framework for Adaptive Probability Thresholding
The rapid growth of video data necessitates efficient summarization techniques to extract critical insights while minimizing computational overhead. This study investigates three pre-trained CNN models—AlexNet (61 M parameters), GoogLeNet (6.8M), and SqueezeNet (1.24M)—with respective inference speeds of 23.8 ms/frame, 18.4 ms/frame, and 12.6 ms/frame. These computational differences significantly affect real-time summarization performance. A probability-driven thresholding mechanism was introduced to optimize object detection, effectively reducing false positives and enhancing classification accuracy. Among the models, GoogLeNet achieved the best trade-off between accuracy (98.96%) and computational efficiency, while AlexNet attained the highest detection accuracy (99.12%) but exhibited a higher false detection rate (5.6%). SqueezeNet, though lightweight and resource-efficient, reached an accuracy of 94.26% but faced limitations in fine-grained recognition tasks. Probability versus correct detection analysis indicated that a threshold of 0.7 maximized accuracy while reducing misclassifications. The proposed adaptive thresholding framework demonstrates strong applicability in real-world scenarios such as wildlife monitoring, intelligent surveillance, and IoT-enabled video analytics. By bridging CNN-based architectures with scalable IoT-driven solutions, this research advances the development of robust, accurate, and efficient video summarization systems for next-generation intelligent environments.
- Open Access
- 21 November 2025
- Published by Elsevier in SLAS Technology
- Vol. 35, 100371
Single-cell RNA insights and densely scaled vision transformer-based MRI classification for precision brain tumors
- 21 November 2025
- Published by Elsevier in SLAS Technology
- Vol. 35, 100371
Open AccessSingle-cell RNA insights and densely scaled vision transformer-based MRI classification for precision brain tumors
- 23 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Standards Magazine
- Vol. PP (24712825) , 1-7
Behavioural Detection of Ransomware and a Comprehensive Approach to Ransomware Prevention
Ransomware is a big and fast-growing cyber threat and can inflict huge financial and operational loss on all types of industries. It took advantage of the vulnerability of the Microsoft Windows and infected over 200,000 computers globally. The increasing attacks, particularly on sensitive services such as Britain’s National Health Service (NHS), indicate that security against cyber-attacks must be quickly improved. The study also reveals the background of ransomware, showing how it has evolved from simple random attacks to more sophisticated ones that incorporate the use of AI. AI is enhancing the efficiency of ransomware by enabling it to perform real-time analysis, intelligent targeting, and adaptive strategies that bypass conventional security. All these improve the efficiency of ransomware in exploiting human vulnerabilities through social engineering. In response to these new threats, this paper prescribes anticipatory security measures. These include the application of AI-based security solutions, behavior-based anomaly detection, and improved employee training and awareness. From the analysis of the relationship between ransomware and artificial intelligence, this research presents future possibilities for cyber threats and security measures that organizations must adopt to protect valuable assets.
- 23 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Standards Magazine
- Vol. PP (24712825) , 1-7
Behavioural Detection of Ransomware and a Comprehensive Approach to Ransomware Prevention
Ransomware is a big and fast-growing cyber threat and can inflict huge financial and operational loss on all types of industries. It took advantage of the vulnerability of the Microsoft Windows and infected over 200,000 computers globally. The increasing attacks, particularly on sensitive services such as Britain’s National Health Service (NHS), indicate that security against cyber-attacks must be quickly improved. The study also reveals the background of ransomware, showing how it has evolved from simple random attacks to more sophisticated ones that incorporate the use of AI. AI is enhancing the efficiency of ransomware by enabling it to perform real-time analysis, intelligent targeting, and adaptive strategies that bypass conventional security. All these improve the efficiency of ransomware in exploiting human vulnerabilities through social engineering. In response to these new threats, this paper prescribes anticipatory security measures. These include the application of AI-based security solutions, behavior-based anomaly detection, and improved employee training and awareness. From the analysis of the relationship between ransomware and artificial intelligence, this research presents future possibilities for cyber threats and security measures that organizations must adopt to protect valuable assets.
- 22 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71 (4) , 11026-11033
Hierarchically Scaled Vision Transformer (HSViT) With Digital Twinned Non-Invasive Neurotherapy to Detect Brain Tumor
Reliable brain tumor analysis from limited MRI data requires lightweight models that provides both accurate classification and clinically useful localization. Conventional Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) struggle in low-data medical settings and are typically limited to single-task classification. We present Hierarchically Scaled Vision Transformer (HSViT), a lightweight transformer for brain MRI that integrates a PreEncoder for edge enhancement, Cross-Layer Attention Refinement (CLAR), a Progressive Token Filter, and a Tumor Context Module (TCM) for efficient tumor classification and localization. HSViT jointly outputs tumor class labels and Region of Interest (ROI) maps, offering diagnostic accuracy with spatial localization. These outputs can also support patient specific digital twin pipelines for non-invasive neurotherapy planning with Transcranial Magnetic Stimulation (TMS) and Transcranial Focused Ultrasound Stimulation (tFUS). On the FigShare dataset, HSViT achieves 97.83% accuracy with precision 99.11%, recall 98.66%, and F1-score 98.88%, with class-wise F1 of 98.47% (Glioma), 98.14% (Meningioma), and 99.13% (Pituitary). Ablation results shows that CLAR improves cross-layer cue stability, the Progressive Token Filter suppresses non informative regions, and TCM sharpens lesion boundaries. The lightweight footprint and token efficiency makes HSViT practical for deployment on resource constrained platforms. Code, training configurations, deterministic environment files, and pre-trained weights are available at: https://github.com/ChauhanPratik/HSViT.
- 22 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71 (4) , 11026-11033
Hierarchically Scaled Vision Transformer (HSViT) With Digital Twinned Non-Invasive Neurotherapy to Detect Brain Tumor
Reliable brain tumor analysis from limited MRI data requires lightweight models that provides both accurate classification and clinically useful localization. Conventional Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) struggle in low-data medical settings and are typically limited to single-task classification. We present Hierarchically Scaled Vision Transformer (HSViT), a lightweight transformer for brain MRI that integrates a PreEncoder for edge enhancement, Cross-Layer Attention Refinement (CLAR), a Progressive Token Filter, and a Tumor Context Module (TCM) for efficient tumor classification and localization. HSViT jointly outputs tumor class labels and Region of Interest (ROI) maps, offering diagnostic accuracy with spatial localization. These outputs can also support patient specific digital twin pipelines for non-invasive neurotherapy planning with Transcranial Magnetic Stimulation (TMS) and Transcranial Focused Ultrasound Stimulation (tFUS). On the FigShare dataset, HSViT achieves 97.83% accuracy with precision 99.11%, recall 98.66%, and F1-score 98.88%, with class-wise F1 of 98.47% (Glioma), 98.14% (Meningioma), and 99.13% (Pituitary). Ablation results shows that CLAR improves cross-layer cue stability, the Progressive Token Filter suppresses non informative regions, and TCM sharpens lesion boundaries. The lightweight footprint and token efficiency makes HSViT practical for deployment on resource constrained platforms. Code, training configurations, deterministic environment files, and pre-trained weights are available at: https://github.com/ChauhanPratik/HSViT.
- Top Cited PapersOpen Access
- 10 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 57705-57719
A Hybrid Brain Tumor Classification Using FL With FedAvg and FedProx for Privacy and Robustness Across Heterogeneous Data Sources
Data privacy and heterogeneity among healthcare settings present fundamental challenges to machine learning (ML) brain tumor classification (BTC) model development based on local data. In this paper, we outline the need to develop an effective brain tumor diagnosis model while ensuring data security. We propose a distributed learning framework based on Federated Learning (FL) to build such a system. This framework combines Federated Averaging (FedAvg) and Federated Proximal (FedProx) to train Convolutional Neural Networks (CNN) on data hosted by multiple clients. FedAvg saves these locally trained models from the patients to make a global model while keeping sensitive patient data decentralized and private. Nevertheless, traditional FL suffers from model instability and divergence due to the disparity of data distributions among clients. FedProx addresses this challenge by introducing a proximal regularization term that penalizes divergence from the global model to stabilize local training considering data heterogeneity. In this work, we utilized the Brain Tumor MRI Dataset from Kaggle and trained the model to classify four brain tumors (BT), namely glioma, meningioma, pituitary tumor, and no-tumor class. Experiments were conducted for ten communication rounds, with each client training the model independently based on its local dataset and without sharing data. Results revealed that the method consistently outperformed accuracy, recall and F1-score, further validating the capabilities of the proposed framework in effectively handling data heterogeneity while conserving data privacy. The global model outperformed models that had been centralized and trained on a smaller, common dataset observed in FL scenarios to improve the potential for collaborative medical diagnosis. This method is especially suitable for medical applications with strict privacy regulations, as it provides an approach to linking heterogeneous data from multiple healthcare institutions.
- 10 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 57705-57719
Top Cited PapersOpen AccessA Hybrid Brain Tumor Classification Using FL With FedAvg and FedProx for Privacy and Robustness Across Heterogeneous Data Sources
Data privacy and heterogeneity among healthcare settings present fundamental challenges to machine learning (ML) brain tumor classification (BTC) model development based on local data. In this paper, we outline the need to develop an effective brain tumor diagnosis model while ensuring data security. We propose a distributed learning framework based on Federated Learning (FL) to build such a system. This framework combines Federated Averaging (FedAvg) and Federated Proximal (FedProx) to train Convolutional Neural Networks (CNN) on data hosted by multiple clients. FedAvg saves these locally trained models from the patients to make a global model while keeping sensitive patient data decentralized and private. Nevertheless, traditional FL suffers from model instability and divergence due to the disparity of data distributions among clients. FedProx addresses this challenge by introducing a proximal regularization term that penalizes divergence from the global model to stabilize local training considering data heterogeneity. In this work, we utilized the Brain Tumor MRI Dataset from Kaggle and trained the model to classify four brain tumors (BT), namely glioma, meningioma, pituitary tumor, and no-tumor class. Experiments were conducted for ten communication rounds, with each client training the model independently based on its local dataset and without sharing data. Results revealed that the method consistently outperformed accuracy, recall and F1-score, further validating the capabilities of the proposed framework in effectively handling data heterogeneity while conserving data privacy. The global model outperformed models that had been centralized and trained on a smaller, common dataset observed in FL scenarios to improve the potential for collaborative medical diagnosis. This method is especially suitable for medical applications with strict privacy regulations, as it provides an approach to linking heterogeneous data from multiple healthcare institutions.
- Open Access
- 23 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 13015-13029
PBVit: A Patch-Based Vision Transformer for Enhanced Brain Tumor Detection
Brain Tumor holds a significant holds in human health, classified into three primary types: glioma, meningioma, and pituitary tumors. Early detection and accurate classification are vital for effective diagnosis and lowering healthcare costs. In PBvit we presents a novel brain tumor detection framework, the Patch Base Vision Transformer (PBVit). PBVit adopts a patch-based approach where input tumor images are divided into fixed-size patches, with each patch treated as a token. These image patches are linearly projected into lower-dimensional token embeddings, and positional encodings are added to help the model understand spatial relationships within the image. PBVit enhances the detection of intricate patterns and anomalies in brain scans, improving diagnostic accuracy. We trained PBVit using the Figshare brain tumor dataset and observed notable performance improvements compared to traditional CNN-based models. The PBVit reached an accuracy of 95.8%, a precision of 95.3%, a recall of 93.2%, and an F1-score of 92%, indicating its robustness in identifying brain tumors. The promising results demonstrate that PBVit can play a important role in facilitating early-stage diagnosis, reducing unnecessary biopsies, and ultimately enhancing patient care, while also showcasing the potential of transformer-based architectures in medical imaging.
- 23 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 13015-13029
Open AccessPBVit: A Patch-Based Vision Transformer for Enhanced Brain Tumor Detection
Brain Tumor holds a significant holds in human health, classified into three primary types: glioma, meningioma, and pituitary tumors. Early detection and accurate classification are vital for effective diagnosis and lowering healthcare costs. In PBvit we presents a novel brain tumor detection framework, the Patch Base Vision Transformer (PBVit). PBVit adopts a patch-based approach where input tumor images are divided into fixed-size patches, with each patch treated as a token. These image patches are linearly projected into lower-dimensional token embeddings, and positional encodings are added to help the model understand spatial relationships within the image. PBVit enhances the detection of intricate patterns and anomalies in brain scans, improving diagnostic accuracy. We trained PBVit using the Figshare brain tumor dataset and observed notable performance improvements compared to traditional CNN-based models. The PBVit reached an accuracy of 95.8%, a precision of 95.3%, a recall of 93.2%, and an F1-score of 92%, indicating its robustness in identifying brain tumors. The promising results demonstrate that PBVit can play a important role in facilitating early-stage diagnosis, reducing unnecessary biopsies, and ultimately enhancing patient care, while also showcasing the potential of transformer-based architectures in medical imaging.
- Open Access
- 13 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 136389-136407
Analyzing Brain Tumor Classification Techniques: A Comprehensive Survey
Classification of brain tumour is critical in medical image analysis and diagnosis as it aids in determining the affinity of cases to select treatment paths. This feature underscores the importance of accurate classification of brain tumours to predict their behavior, determine suitable management methods, and thus improve the performance of patients. The last few years have revolutionized medical picture grouping and brain tumour studies, thanks to machine learning tools, primarily CNNs. They demonstrate much potential for nearly extracting all features of interest from medical images and using them to classify tumours with high levels of accuracy. CNNs have complied with high accuracy rates, especially compared to the other forms of machine learning in this field. Due to the inherent ability to extract various aspects of the image along with the hierarchical detail of the picture, they are more suitable for use in operations like brain tumour classification. Similarly, the advancements in the field of natural language processing with transformer models have resulted in increased interest in the practical application of these types of models in image classification tasks. However, the application of ML models in the classification of brain tumours is still limited, therefore requiring continued research in this area. Therefore, the question about CNNs and Transformers’ performance differences within the framework of the brain tumour classification becomes relatively more significant. This survey aims to provide recent contributions of related work in the domain and highlight the dynamism that has emerged due to such state-of-the-art methods.
- 13 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 136389-136407
Open AccessAnalyzing Brain Tumor Classification Techniques: A Comprehensive Survey
Classification of brain tumour is critical in medical image analysis and diagnosis as it aids in determining the affinity of cases to select treatment paths. This feature underscores the importance of accurate classification of brain tumours to predict their behavior, determine suitable management methods, and thus improve the performance of patients. The last few years have revolutionized medical picture grouping and brain tumour studies, thanks to machine learning tools, primarily CNNs. They demonstrate much potential for nearly extracting all features of interest from medical images and using them to classify tumours with high levels of accuracy. CNNs have complied with high accuracy rates, especially compared to the other forms of machine learning in this field. Due to the inherent ability to extract various aspects of the image along with the hierarchical detail of the picture, they are more suitable for use in operations like brain tumour classification. Similarly, the advancements in the field of natural language processing with transformer models have resulted in increased interest in the practical application of these types of models in image classification tasks. However, the application of ML models in the classification of brain tumours is still limited, therefore requiring continued research in this area. Therefore, the question about CNNs and Transformers’ performance differences within the framework of the brain tumour classification becomes relatively more significant. This survey aims to provide recent contributions of related work in the domain and highlight the dynamism that has emerged due to such state-of-the-art methods.
- 20 June 2024
- Published by Taylor & Francis
- p. 82-98
Big Data Forensics
The volume of digital data is now rising at an almost unfathomable rate because of the computational and storage power of the information age. There are now more chances than ever before, thanks to the big data (BD) revolution. BD is a vital resource for digital forensic investigation. Large datasets are readily available, which can be careful for data integrity and helpful in some situations but also creates new challenges for study. Existing open source and paid infrastructures and tools have a restricted range of applications and fall short of the anticipated turnaround times for large dataset analysis. Cyber forensic investigators struggle to find the pertinent evidence in a large dataset, gather it, analyze it, and recreate historical events. In order to examine the issues in the digital forensic paradigm, with this work, we introduce the concept of BD Forensics and critically evaluate forensics of the BD area. We suggest a conceptual framework to link large data forensics inquiries and offer fresh perspectives on examining criminal situations.
- 20 June 2024
- Published by Taylor & Francis
- p. 82-98
Big Data Forensics
The volume of digital data is now rising at an almost unfathomable rate because of the computational and storage power of the information age. There are now more chances than ever before, thanks to the big data (BD) revolution. BD is a vital resource for digital forensic investigation. Large datasets are readily available, which can be careful for data integrity and helpful in some situations but also creates new challenges for study. Existing open source and paid infrastructures and tools have a restricted range of applications and fall short of the anticipated turnaround times for large dataset analysis. Cyber forensic investigators struggle to find the pertinent evidence in a large dataset, gather it, analyze it, and recreate historical events. In order to examine the issues in the digital forensic paradigm, with this work, we introduce the concept of BD Forensics and critically evaluate forensics of the BD area. We suggest a conceptual framework to link large data forensics inquiries and offer fresh perspectives on examining criminal situations.