Deepsubhra Guha Roy
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
Deepsubhra Guha Roy has a total of 49 co-authors. Here are the top co-authors by number of shared publications:
- Debashis De
Department of Computer Science and Engineering, Maulana Abul Kalam Azad University of Technology, West Bengal, Kolkata, India
5 shared publications - Piyali Datta
University of Engineering and Management
5 shared publications - Moutushi Singh
University of Engineering and Management
4 shared publications - Bipasha Mahato
IEM Centre of Excellence for Cloud Computing & IoT, Department of CSE(AIML), Institute of Engineering & Management, University of Engineering and Management, Kolkata, India
3 shared publications Piyali Datta
Institute of Engineering & Management, University of Engineering & Management
3 shared publications- Puja Das
Department of CSE-CPS, Jain University, Kanakpura, Bengaluru
3 shared publications
Latest Publications
- Open Access
- 29 September 2026
- Published by Springer Nature in Discover Networks
- Vol. 2 (1) , 37
Biofeedback-augmented LLM framework for personalized dietary risk assessment
- 29 September 2026
- Published by Springer Nature in Discover Networks
- Vol. 2 (1) , 37
Open AccessBiofeedback-augmented LLM framework for personalized dietary risk assessment
- 2 September 2026
- Published by Springer Nature
- p. 163-174
MedAxis: Gen-AI Framework for Real-Time Triage and Crisis Management in Smart Cities
- 2 September 2026
- Published by Springer Nature
- p. 163-174
MedAxis: Gen-AI Framework for Real-Time Triage and Crisis Management in Smart Cities
- 6 January 2026
- Published by Springer Nature
- p. 207-219
Stocks in Sync: Cluster-Aware Deep Learning for Multi-stock Forecasting in Financial Market
- 6 January 2026
- Published by Springer Nature
- p. 207-219
Stocks in Sync: Cluster-Aware Deep Learning for Multi-stock Forecasting in Financial Market
- 2 January 2026
- Published by Springer Nature
- p. 425-435
Harnessing the Power of LSTM, ARIMA, and Dense Models for Accurate Temperature Prediction with Time Series Analysis
- 2 January 2026
- Published by Springer Nature
- p. 425-435
Harnessing the Power of LSTM, ARIMA, and Dense Models for Accurate Temperature Prediction with Time Series Analysis
- 2 January 2026
- Published by Springer Nature
- p. 71-82
NooGenie: Unlocking Noonan Syndrome with Llama 3
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (1) , 2007-2015
Proactive Zero-Trust Intrusion Detection for Consumer IoT Applications Using Lightweight Ensemble Learning With Anomaly Analysis
The rapid proliferation of consumer IoT devices from smart home hubs to wearables has expanded the attack surface, introducing new security and privacy challenges. Traditional Intrusion Detection Systems (IDS) often rely on implicit trust and heavyweight computation, making them unsuitable for resource-constrained consumer electronics. This paper presents a proactive, lightweight zero-trust IDS tailored for consumer applications. Our two-layer architecture integrates a supervised stacked ensemble classifier (Random Forest, XGBoost, Light-GBM) to detect known threats and an unsupervised DBSCAN (Density-Based Spatial Clustering of Applications with Noise)-based anomaly detector to identify zero-day attacks. We introduce a feature reduction pipeline driven by correlation and variance analysis, reducing the feature set by over 50% to fit edge hardware constraints. Evaluated on the CICIDS collection dataset (over 9 million flows), the framework achieves 98.48% accuracy while maintaining real-time processing capability on Raspberry Pi-class hardware. By continuously scrutinizing both malicious and benign traffic, our system delivers proactive, trust-enhancing defense critical to modern consumer IoT applications.
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (1) , 2007-2015
Proactive Zero-Trust Intrusion Detection for Consumer IoT Applications Using Lightweight Ensemble Learning With Anomaly Analysis
The rapid proliferation of consumer IoT devices from smart home hubs to wearables has expanded the attack surface, introducing new security and privacy challenges. Traditional Intrusion Detection Systems (IDS) often rely on implicit trust and heavyweight computation, making them unsuitable for resource-constrained consumer electronics. This paper presents a proactive, lightweight zero-trust IDS tailored for consumer applications. Our two-layer architecture integrates a supervised stacked ensemble classifier (Random Forest, XGBoost, Light-GBM) to detect known threats and an unsupervised DBSCAN (Density-Based Spatial Clustering of Applications with Noise)-based anomaly detector to identify zero-day attacks. We introduce a feature reduction pipeline driven by correlation and variance analysis, reducing the feature set by over 50% to fit edge hardware constraints. Evaluated on the CICIDS collection dataset (over 9 million flows), the framework achieves 98.48% accuracy while maintaining real-time processing capability on Raspberry Pi-class hardware. By continuously scrutinizing both malicious and benign traffic, our system delivers proactive, trust-enhancing defense critical to modern consumer IoT applications.
- 31 August 2025
- Published by Springer Nature
- p. 429-441
MedMate: A Contextual Approach for Disease Diagnosis Using Retrieval-Augmented Generation
- 31 August 2025
- Published by Springer Nature
- p. 429-441
MedMate: A Contextual Approach for Disease Diagnosis Using Retrieval-Augmented Generation
- 21 August 2025
- Published by Springer Nature in Theoretical and Applied Climatology
- Vol. 156 (9) , 1-18
STAT-X: short-term atmospheric temperature forecasting using machine learning models with explainable-AI
- 21 August 2025
- Published by Springer Nature in Theoretical and Applied Climatology
- Vol. 156 (9) , 1-18
STAT-X: short-term atmospheric temperature forecasting using machine learning models with explainable-AI
- Open Access
- 1 January 2025
- Published by Wiley in International Journal of Distributed Sensor Networks
- Vol. 2025 (1)
Quality of Experience Aware Service Selection Model to Empower Edge Computing in IoT
Quality of experience‐aware service selection can significantly remove well‐known scalability issues of an Internet of Things (IoT) architecture. In traditional IoT architecture, several heterogeneous data streams from connected nodes are transmitted through gateways to the remote mobile cloud servers. The entire procedure is time‐ and energy‐consuming if the target dataset is comparatively small and uninterrupted. Also, using this conventional technique, the reliability grade drops significantly to meet additional security‐related quality of service (QoS) requirements compared to the service cost. We propose a quality of experience‐aware task rescheduling model using edge modules that offer territory‐based three‐layered edge IoT data analysis and service selection. The observation module at the application layer takes a near‐optimal remark upon each usage metric having distinct QoS components. Meanwhile, the QoS manager at the network layer handles network traffic due to the load associated with heterogeneous service needs. The precision of the knowledge is assured to the service manager through the sensing layer with few adaptability characteristics towards assorted service requests. The proposed three‐layered energy‐efficient model helps minimize data delivery time with minimal cost and optimized quality assurance for service‐based IoT infrastructures like smart agriculture, patient monitoring, and student monitoring.
- 1 January 2025
- Published by Wiley in International Journal of Distributed Sensor Networks
- Vol. 2025 (1)
Open AccessQuality of Experience Aware Service Selection Model to Empower Edge Computing in IoT
Quality of experience‐aware service selection can significantly remove well‐known scalability issues of an Internet of Things (IoT) architecture. In traditional IoT architecture, several heterogeneous data streams from connected nodes are transmitted through gateways to the remote mobile cloud servers. The entire procedure is time‐ and energy‐consuming if the target dataset is comparatively small and uninterrupted. Also, using this conventional technique, the reliability grade drops significantly to meet additional security‐related quality of service (QoS) requirements compared to the service cost. We propose a quality of experience‐aware task rescheduling model using edge modules that offer territory‐based three‐layered edge IoT data analysis and service selection. The observation module at the application layer takes a near‐optimal remark upon each usage metric having distinct QoS components. Meanwhile, the QoS manager at the network layer handles network traffic due to the load associated with heterogeneous service needs. The precision of the knowledge is assured to the service manager through the sensing layer with few adaptability characteristics towards assorted service requests. The proposed three‐layered energy‐efficient model helps minimize data delivery time with minimal cost and optimized quality assurance for service‐based IoT infrastructures like smart agriculture, patient monitoring, and student monitoring.
- 23 December 2024
- Published by Springer Nature
- p. 129-143
An Overview of the Discrete Logarithm Problem in Cryptography
- 23 December 2024
- Published by Springer Nature
- p. 129-143
An Overview of the Discrete Logarithm Problem in Cryptography