Sunil Kumar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sunil Kumar has a total of 79 co-authors. Here are the top co-authors by number of shared publications:
- Megha Jain
School of Computing Science and Engineering, VIT Bhopal University, Sehore, India
3 shared publications - Ravi Verma
School of Computing Science and Engineering, VIT Bhopal University, Sehore, India
3 shared publications Ananya Agrawal
Manipal University Jaipur
2 shared publicationsDilip Kumar
Department of Computer Science and Engineering National Institute of Technology Jamshedpur India
2 shared publicationsGaurav Thakur
Department of Computer Science and Informatics, Central University of Himachal Pradesh, Dharamshala, India
2 shared publications- Gyanendra Kumar
Manipal University Jaipur
2 shared publications
Latest Publications
- Open Access
- 8 September 2026
- Published by Springer Nature in Discover Artificial Intelligence
- Vol. 6  (1) , 1090
Transformer-based deep learning models for lung cancer detection and classification: a comprehensive review
Lung cancer is one of the most aggressive and life-threatening diseases worldwide, responsible for a large percentage of cancer-related deaths each year. Early and accurate diagnosis plays a crucial role in improving patient outcomes and reducing mortality rates. However, traditional diagnostic approaches often fail to detect tumors in their initial stages, leading to delayed treatment. Recent developments in deep learning have introduced transformer-based models that demonstrate remarkable capability in analyzing medical images. These models effectively capture spatial and contextual information from medical imaging, enabling better detection, segmentation, and classification of lung tumors. This review presents a detailed examination of transformer-based architectures applied to lung cancer detection and classification. The review summarizes findings from recently published studies to examine current transformer-based approaches, their applications, and emerging research trends. It discusses their performance, strengths, and limitations compared with earlier methods and explores key challenges such as limited labeled data, dataset imbalance, computational complexity, and lack of interpretability. The reviewed studies show that transformer-based models deliver promising performance, although their results vary depending on the dataset, validation method, and model design. The paper also highlights potential future directions, including model optimization, data standardization, and validation across diverse clinical environments, to enhance the reliability and integration of transformer-based models in healthcare applications.
- 8 September 2026
- Published by Springer Nature in Discover Artificial Intelligence
- Vol. 6  (1) , 1090
Open AccessTransformer-based deep learning models for lung cancer detection and classification: a comprehensive review
Lung cancer is one of the most aggressive and life-threatening diseases worldwide, responsible for a large percentage of cancer-related deaths each year. Early and accurate diagnosis plays a crucial role in improving patient outcomes and reducing mortality rates. However, traditional diagnostic approaches often fail to detect tumors in their initial stages, leading to delayed treatment. Recent developments in deep learning have introduced transformer-based models that demonstrate remarkable capability in analyzing medical images. These models effectively capture spatial and contextual information from medical imaging, enabling better detection, segmentation, and classification of lung tumors. This review presents a detailed examination of transformer-based architectures applied to lung cancer detection and classification. The review summarizes findings from recently published studies to examine current transformer-based approaches, their applications, and emerging research trends. It discusses their performance, strengths, and limitations compared with earlier methods and explores key challenges such as limited labeled data, dataset imbalance, computational complexity, and lack of interpretability. The reviewed studies show that transformer-based models deliver promising performance, although their results vary depending on the dataset, validation method, and model design. The paper also highlights potential future directions, including model optimization, data standardization, and validation across diverse clinical environments, to enhance the reliability and integration of transformer-based models in healthcare applications.
- Open Access
- 13 August 2026
- Published by Springer Nature in Discover Computing
- Vol. 29  (1) , 533
A consent-based medical data sharing and edge offloading scheme based on blockchain and deep reinforcement learning
In the current digital era, the storage of electronic health records on centralized platforms presents significant integrity, privacy and security challenges. Further, access to this stored healthcare data should be quick and efficient, especially during emergencies. Blockchain and edge computing brought a great revolution in managing healthcare data by ensuring security, immutability, and decentralized data sharing with reduced latency. But, the integration of edge computing with the blockchain networks is still a gap to achieve ideal healthcare goals of data security with real-time data processing. The contribution of this work is two-fold. First, a novel deep reinforcement learning based medical data offloading scheme is proposed for offloading healthcare data to the nearby edge servers from the end users. The learning policy uses the proximal policy optimization algorithm for making the optimal offloading decision and minimizes the overall delay and energy consumption of healthcare devices and edge servers. Second, we proposed a secure, scalable, and consent-based data sharing scheme among multiple stakeholders such as patients, hospitals, doctors, healthcare research institutes etc. The EHR sharing scheme uses the AES and RSA algorithms for encryption, which ensures only authorized and consent-based access to the sensitive data stored in IPFS. The performance of the proposed offloading scheme is evaluated in terms of delay and energy consumption whereas data sharing scheme is evaluated in terms of latency and throughput using Hyperledger Besu and Hyperledger Caliper platforms. The experimental study exhibits that the proposed approach is both feasible and scalable, making it suitable for integration into the e-healthcare systems.
- 13 August 2026
- Published by Springer Nature in Discover Computing
- Vol. 29  (1) , 533
Open AccessA consent-based medical data sharing and edge offloading scheme based on blockchain and deep reinforcement learning
In the current digital era, the storage of electronic health records on centralized platforms presents significant integrity, privacy and security challenges. Further, access to this stored healthcare data should be quick and efficient, especially during emergencies. Blockchain and edge computing brought a great revolution in managing healthcare data by ensuring security, immutability, and decentralized data sharing with reduced latency. But, the integration of edge computing with the blockchain networks is still a gap to achieve ideal healthcare goals of data security with real-time data processing. The contribution of this work is two-fold. First, a novel deep reinforcement learning based medical data offloading scheme is proposed for offloading healthcare data to the nearby edge servers from the end users. The learning policy uses the proximal policy optimization algorithm for making the optimal offloading decision and minimizes the overall delay and energy consumption of healthcare devices and edge servers. Second, we proposed a secure, scalable, and consent-based data sharing scheme among multiple stakeholders such as patients, hospitals, doctors, healthcare research institutes etc. The EHR sharing scheme uses the AES and RSA algorithms for encryption, which ensures only authorized and consent-based access to the sensitive data stored in IPFS. The performance of the proposed offloading scheme is evaluated in terms of delay and energy consumption whereas data sharing scheme is evaluated in terms of latency and throughput using Hyperledger Besu and Hyperledger Caliper platforms. The experimental study exhibits that the proposed approach is both feasible and scalable, making it suitable for integration into the e-healthcare systems.
- Open Access
- 8 August 2026
- Published by Springer Nature in Discover Applied Sciences
- Vol. 8  (11)
Comparing various vision transformer architectures for lung cancer classification
This study presents a comprehensive evaluation of vision transformer-based architectures for lung histopathology image classification using the LC25000 dataset. We systematically compare custom-trained transformer models and ImageNet-pretrained architectures, including Broad Vision Transformer, DeiT, Swin Transformer, and ConvNeXt, under a unified experimental protocol. The results demonstrate that pretrained models significantly outperform models trained from scratch, achieving up to 99.7% classification accuracy and consistently higher precision, recall, and F1-scores across all diagnostic categories. Detailed analysis reveals that transfer learning enables more robust capture of tissue-level structural patterns, particularly in distinguishing adenocarcinoma and squamous cell carcinoma. The findings highlight the importance of strong initialization for transformer-based medical imaging models and provide practical insights for selecting architectures in data-limited histopathology applications.
- 8 August 2026
- Published by Springer Nature in Discover Applied Sciences
- Vol. 8  (11)
Open AccessComparing various vision transformer architectures for lung cancer classification
This study presents a comprehensive evaluation of vision transformer-based architectures for lung histopathology image classification using the LC25000 dataset. We systematically compare custom-trained transformer models and ImageNet-pretrained architectures, including Broad Vision Transformer, DeiT, Swin Transformer, and ConvNeXt, under a unified experimental protocol. The results demonstrate that pretrained models significantly outperform models trained from scratch, achieving up to 99.7% classification accuracy and consistently higher precision, recall, and F1-scores across all diagnostic categories. Detailed analysis reveals that transfer learning enables more robust capture of tissue-level structural patterns, particularly in distinguishing adenocarcinoma and squamous cell carcinoma. The findings highlight the importance of strong initialization for transformer-based medical imaging models and provide practical insights for selecting architectures in data-limited histopathology applications.
- 2 July 2026
- Published by Springer Nature
- p. 214-224
Security Vulnerabilities in AI Assistants and Their Implications for IoT-Based Environmental Systems
- 2 July 2026
- Published by Springer Nature
- p. 214-224
Security Vulnerabilities in AI Assistants and Their Implications for IoT-Based Environmental Systems
- Open Access
- 26 June 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16  (1)
A secure and lightweight cryptographic-machine learning framework for IoT-based cyber defense in resource-constrained environments
The rapid expansion of Internet of Things (IoT) systems has introduced significant security challenges, particularly in resource-constrained environments where traditional security mechanisms are often impractical. This paper presents a secure and lightweight hybrid framework that integrates cryptographic techniques with machine learning–based anomaly detection for IoT-based cyber defense. The proposed framework employs Elliptic Curve Cryptography (ECC) for key exchange, SPECK for lightweight encryption, and SHA-3 for data integrity, combined with a Random Forest classifier for anomaly detection. The framework is implemented and evaluated on a Raspberry Pi–based edge environment using the CIC-BCCC-NRC-IoT-2023 dataset. Experimental results demonstrate an accuracy of 89.5% and an F1-score of 90%, with an average end-to-end latency of 1.08 ms and energy consumption of approximately 4.5 mJ per inference. These results indicate that the proposed approach achieves a practical balance between security, computational efficiency, and detection performance under constrained conditions. While the framework shows promising results, its evaluation is limited to a controlled setup and a single primary dataset. Future work will focus on cross-dataset validation, adversarial robustness, and large-scale deployment analysis.
- 26 June 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16  (1)
Open AccessA secure and lightweight cryptographic-machine learning framework for IoT-based cyber defense in resource-constrained environments
The rapid expansion of Internet of Things (IoT) systems has introduced significant security challenges, particularly in resource-constrained environments where traditional security mechanisms are often impractical. This paper presents a secure and lightweight hybrid framework that integrates cryptographic techniques with machine learning–based anomaly detection for IoT-based cyber defense. The proposed framework employs Elliptic Curve Cryptography (ECC) for key exchange, SPECK for lightweight encryption, and SHA-3 for data integrity, combined with a Random Forest classifier for anomaly detection. The framework is implemented and evaluated on a Raspberry Pi–based edge environment using the CIC-BCCC-NRC-IoT-2023 dataset. Experimental results demonstrate an accuracy of 89.5% and an F1-score of 90%, with an average end-to-end latency of 1.08 ms and energy consumption of approximately 4.5 mJ per inference. These results indicate that the proposed approach achieves a practical balance between security, computational efficiency, and detection performance under constrained conditions. While the framework shows promising results, its evaluation is limited to a controlled setup and a single primary dataset. Future work will focus on cross-dataset validation, adversarial robustness, and large-scale deployment analysis.
- 4 June 2026
- Published by Springer Nature
- p. 85-97
A Comparative Analysis of Cryptographic Algorithms in IoT Environment
- 4 June 2026
- Published by Springer Nature
- p. 85-97
A Comparative Analysis of Cryptographic Algorithms in IoT Environment
- Open Access
- 4 April 2026
- Published by Springer Nature in Discover Artificial Intelligence
- Vol. 6  (1)
Privacy- preserving network slice classification in 6G using federated CNN-BiLSTM
This paper presents a federated learning (FL) framework that integrates a lightweight Convolutional Neural Network (CNN) with a Bidirectional Long Short-Term Memory (BiLSTM) network for privacy-preserving 6 G network slice classification. Using the simulated 6 G Network Slicing Dataset, the model learns from distributed edge clients without accessing raw user data, where the CNN extracts spatial QoS features and the BiLSTM models temporal dependencies. Federated training is performed using FedAvg, aggregating locally trained client models at the orchestrator. The proposed federated CNN–BiLSTM architecture achieves strong experimental performance- including 97.8% cost efficiency, 97.9% energy efficiency, 1.3 ms jitter, 0.22 s latency, 98.6% QoS score, 0.045 BER, 1.6% packet drop rate, and 960 Mbps throughput, outperforming baselines such as FDRL-RAN, DRL-DDQT, and centralized CNN–BiLSTM. Although evaluated on synthetic data, the model shows high potential for real-world 6 G deployment. Future work includes integrating differential privacy and secure aggregation. Unlike prior FL methods that rely on reinforcement learning or heavy optimization-based slice controllers, the proposed design leverages a compact 145k-parameter spatio-temporal model ( $$<8$$ MB) tailored for low-power 6 G edge devices, offering an efficient and privacy-aware solution for next-generation slice classification.- 4 April 2026
- Published by Springer Nature in Discover Artificial Intelligence
- Vol. 6  (1)
Open AccessPrivacy- preserving network slice classification in 6G using federated CNN-BiLSTM
This paper presents a federated learning (FL) framework that integrates a lightweight Convolutional Neural Network (CNN) with a Bidirectional Long Short-Term Memory (BiLSTM) network for privacy-preserving 6 G network slice classification. Using the simulated 6 G Network Slicing Dataset, the model learns from distributed edge clients without accessing raw user data, where the CNN extracts spatial QoS features and the BiLSTM models temporal dependencies. Federated training is performed using FedAvg, aggregating locally trained client models at the orchestrator. The proposed federated CNN–BiLSTM architecture achieves strong experimental performance- including 97.8% cost efficiency, 97.9% energy efficiency, 1.3 ms jitter, 0.22 s latency, 98.6% QoS score, 0.045 BER, 1.6% packet drop rate, and 960 Mbps throughput, outperforming baselines such as FDRL-RAN, DRL-DDQT, and centralized CNN–BiLSTM. Although evaluated on synthetic data, the model shows high potential for real-world 6 G deployment. Future work includes integrating differential privacy and secure aggregation. Unlike prior FL methods that rely on reinforcement learning or heavy optimization-based slice controllers, the proposed design leverages a compact 145k-parameter spatio-temporal model ( $$<8$$ MB) tailored for low-power 6 G edge devices, offering an efficient and privacy-aware solution for next-generation slice classification. - 11 February 2026
- Published by Springer Nature in Quantum Information Processing
- Vol. 25  (2) , 62
Composable security of a hybrid BB84–E91 protocol: toward synchronization-resilient quantum key distribution
- 11 February 2026
- Published by Springer Nature in Quantum Information Processing
- Vol. 25  (2) , 62
Composable security of a hybrid BB84–E91 protocol: toward synchronization-resilient quantum key distribution
- Open Access
- 1 February 2026
- Published by Springer Nature in Discover Computing
- Vol. 29  (1) , 66
Binary classification of lung cancer using vision transformer models on CT images
- 1 February 2026
- Published by Springer Nature in Discover Computing
- Vol. 29  (1) , 66
Open AccessBinary classification of lung cancer using vision transformer models on CT images