Spatio-Temporal Polyp Detection in Colonoscopy Videos Using EfficientNet-B0 and Temporal Transformer Advait Sahadev, Tharun H, Ayush Tiwari, S. Murugaveni 2026 International Conference on Recent Advances in Electrical Electronics Ubiquitous Communication and Computational Intelligence Raeeucci 2026, 2026 In this paper, a multi-task deep learning approach is proposed for video based polyp detection and risk assessment in colonoscopy videos. The proposed model combines the EfficientNet-B0 backbone to extract Spatial features and task-specific heads for bounding box regression and lesion risk classification. The backbone is initialized with ImageNet-pretrained weights and kept fixed to guarantee the stability of feature transfer, while using lightweight trainable heads for cost-effective adaptation to medical data. The model is tested using standard spatial localization metrics like IoU, Dice coefficient, Precision, Recall as well as F1-score over annotated polyp datasets. In order to evaluate real-time effectiveness, the model is also evaluated with temporal stability indicators (center drift, variance of bounding box area and detection continuity) on video sequences. With multi-task learning, the lesion localization and severity estimation tasks can be learned simultaneously, which allows modeling of complex transformations and increases efficiency (the total parameters are about <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{4. 3 4}$</tex> million and that for training is roughly 0.33 million). Experimental results show that Temporal modeling enhance localization stability, reducing central drift by 45.8 % and area variance by over 60.8 %, increasing temporal smoothness.
Deep Learning Models for Land Cover Classification and Change Detection in Coastal Zones S. Revathi, P. Selvakumar, T. C. Manjunath, S. Murugaveni, P. Varun, Bottu Giridharan Sivakumar Earth Observation and Deep Learning for Coastal Monitoring Mapping and Management, 2026 Coastal regions represent some of the most dynamic and ecologically complex areas on Earth, where terrestrial and marine systems interact in intricate ways, creating unique landscapes that are both highly productive and highly sensitive to environmental changes. The classification of land cover in these regions is inherently challenging due to the spatial heterogeneity, temporal variability, and fine-scale interactions between natural and anthropogenic elements. Unlike more homogenous inland areas, coasts are characterized by mosaics of sandy beaches, rocky shores, mudflats, estuaries, wetlands, mangroves, saltmarshes, and urban developments, all of which can exhibit significant seasonal and tidal variations. This variability complicates traditional land cover classification methods that often rely on spectral signatures from remote sensing imagery, as the spectral characteristics of coastal features may overlap or shift with changing environmental conditions, tides, or sediment composition.
AI- Powered Malware Detection and Behavioral Analysis P. Selvakumar, Vaishali Rahate, A. S. Deeppana, Yukta Sawalkar, C. John Paul, S. Murugaveni Examining Vulnerabilities and Adversarial Exploitation of AI and Llms, 2026 In the rapidly evolving landscape of cybersecurity, malware continues to pose significant threats to individuals, enterprises, and critical infrastructures. Traditional signature-based detection techniques, though effective against known threats, fall short when confronted with sophisticated, polymorphic, and zero-day malware. This limitation has fueled research into more intelligent, adaptive detection mechanisms that can identify malicious software even when it exhibits novel patterns or obfuscation strategies. Static malware analysis, unlike dynamic analysis, focuses on examining the intrinsic attributes of executable files without executing them, making it safer, faster, and less resource-intensive. Static features typically include opcode sequences, bytecode patterns, control flow graphs, API call frequency distributions, file headers, string literals, and metadata extracted from Portable Executable (PE) files or other binary formats.
AI Security Threats in 5G-Enabled IoT Device Management Kadimisetty Mahendra Kumar, Devagudi Bapuji Reddy, Paidi Rushendra, S. Murugaveni Proceedings of the 6th International Conference on Smart Electronics and Communication Icosec 2025, 2025 The integration of fifth-generation (5G) networks with Internet of Things (IoT) ecosystems significantly enhances connectivity, scalability, and efficiency. However, this interconnection introduces serious cybersecurity challenges, including data breaches, denial-of-service (DoS) attacks, and adversarial manipulation of artificial intelligence (AI)-based systems. The vast number of heterogeneous devices amplifies vulnerabilities in authentication, privacy, and trust management. Existing studies propose blockchain, adaptive machine learning, and hybrid intrusion detection models, but many approaches are limited in scalability, robustness against adversarial AI, and practical deployment in mobile edge computing (MEC). To address these gaps, this work proposes a layered 5G–IoT security framework integrating blockchain-based authentication, AI-driven threat detection, and privacy-preserving input handling. The methodology employs ensemble machine learning classifiers, federated learning for decentralized model updates, and hashed identifiers to ensure privacy without compromising detection accuracy. Comparative evaluation across classifiers demonstrates the superiority of Support Vector Machines (SVM), which achieved 100% accuracy on the test dataset, while logistic regression and gradient boosting underperformed. The novelty of this work lies in the combined use of AI with privacy-preserving mechanisms and federated updates, making it suitable for real-time edge deployments. This study highlights key research gaps, provides a comparative analysis of methods, and discusses limitations, challenges, and future opportunities for securing 5G-enabled IoT ecosystems.
Fetal Health Prediction: Comparative Analysis of Machine Learning Models Sheron S, Nehal Sreejith, Adithya B Chandran, S. Murugaveni 2025 International Conference on Recent Advances in Electrical Electronics Ubiquitous Communication and Computational Intelligence Raeeucci 2025, 2025 Monitoring the health of the fetus is crucial to the mother's and the child's well-being. Even though cardiotocography (CTG) is used to assess fetal conditions, interpretation is usually done manually, which can be inaccurate and error-prone. This study investigates automating the classification of fetal health using machine learning, which would improve the process's efficiency and dependability. To improve predictive performance, we create hybrid models integrating Support Vector Machine (SVM) with Bagging and Multi-Layer Perceptron (MLP) with Bagging which achieved an accuracy of 0.99. Our goal is to correctly identify fetal health as normal, suspicious, or pathological by pre-processing CTG data, choosing important features, and training these models. Our method makes use of ensemble learning strategies to enhance dependability and facilitate clinical judgment. The findings demonstrate how AI- driven techniques can provide quicker and more objective evaluations, helping medical practitioners make early diagnoses and take timely action.
Utilizing Explainable Artificial Intelligence for Parkinson’s Disease Diagnosis through the Analysis of Spiral and Wave Drawings with Integrated Data Augmentation S. Murugaveni, Harisudha Kuresan, N. Sri Sai Charan Reddy, Perala Manoj, Gaddam Rohit Advances in Computational Intelligence for Health Informatics and Computer Aided Diagnosis Methods Applications and Tools, 2025 Parkinson’s disease (PD) is a chronic neurodegenerative condition affecting the central nervous system, resulting in motor function impairment due to dopamine shortage. PD manifests in movement difficulties such as tremors, stiffness, and bradykinesia, impacting precise motor control required for tasks like drawing spirals and waves. The identification of biomarkers associated with health conditions is critical for advancing clinical decision support systems, with PD showing a notable correlation between disease severity and impaired handwriting, along with reduced speed and pressure during sketching or writing. This study prioritizes transparency and reliability to enhance the early diagnostic precision of PD through VGG19 with an attention mechanism. Understanding the decision-making process of classifiers in predicting PD is complex, necessitating explainable artificial intelligence (XAI) for refining clinical procedures and ensuring transparency. Techniques like local interpretable model-agnostic explanation method (LIME) and Shapley additive explanations (SHAP) are employed to identify specific segments in spiral and wave diagrams that contribute significantly to the model’s predictions, revealing the details of how the proposed model works. To address the limited dataset, data augmentation techniques expand the dataset, enhancing the model’s robustness. These approaches provide localized interpretation, offering a clearer understanding of the proposed model’s functioning.
CPW fed ultrawideband antenna with band notched performance for polarization diversity Journal of Advanced Research in Dynamical and Control Systems, 2017
Reduced complexity by incorporating sphere decoder with MIMO STBC HARQ systems International Journal of Control Theory and Applications, 2016
Opportunistic communication between various users using distributed dynamic spectrum protocol International Journal of Applied Engineering Research, 2015
RECENT SCHOLAR PUBLICATIONS
Spatio-Temporal Polyp Detection in Colonoscopy Videos Using EfficientNet-B0 and Temporal Transformer A Sahadev, H Tharun, A Tiwari, S Murugaveni 2026 International Conference on Recent Advances in Electrical, Electronics … , 2026 2026
Fetal Health Prediction: Comparative Analysis of Machine Learning Models S Sheron, N Sreejith, AB Chandran, S Murugaveni 2025 International Conference on Recent Advances in Electrical, Electronics … , 2025 2025 Citations: 1
Utilizing Explainable Artificial Intelligence for Parkinson's Disease Diagnosis through the Analysis of Spiral and Wave Drawings with Integrated Data Augmentation S Murugaveni, H Kuresan, NSSC Reddy, P Manoj, G Rohit Advances in Computational Intelligence for Health Informatics and Computer … , 2025 2025
Deep Learning-Based Multiclass Classification for Dental Disease Detection Using DenseNet-201 AB Chandran, S Sheron, N Sreejith, VM Harshini, S Murugaveni International Conference on Data Science and Management, 215-227 , 2024 2024
Real-Time Detection of Anxiety and Panic Attacks ASA Princella, H Kuresan, M Anand, S Murugaveni International Conference on Data Science and Management, 269-281 , 2024 2024
Evaluating the Effectiveness of Deep Learning Models in Parkinson’s Disease Diagnosis Using Magnetic Resonance Imaging S Balamithra, H Kuresan, S Murugaveni, VM Harshini International Conference on Data Science and Management, 241-253 , 2024 2024
Retraction Note: A novel approach of low complexity distributed UA algorithm is used for traffic load balancing and interference in next generation networks E Elamaran, S Murugaveni, S Jyothi, MR Prabhu, MP Chitra, V Talasila Optical and Quantum Electronics 56 (10), 1679 , 2024 2024
RETRACTED ARTICLE: A novel approach of low complexity distributed UA algorithm is used for traffic load balancing and interference in next generation networks E Elamaran, S Murugaveni, S Jyothi, MR Prabhu, MP Chitra, V Talasila Optical and Quantum Electronics 56 (4), 525 , 2024 2024 Citations: 1
Layering of edge node for jamming attack detection and elimination in wireless sensor network S Murugaveni, B Priyalakshmi Concurrency and Computation: Practice and Experience 35 (22), e7737 , 2023 2023 Citations: 8
Retraction Note to: A novel approach for non-orthogonal multiple access for delay sensitive industrial IoT communications for smart autonomous factories S Murugaveni, K Mahalakshmi Journal of Ambient Intelligence and Humanized Computing 14 (Suppl 1), 151-151 , 2023 2023 Citations: 1
Emperor penguin optimized Q learning method for energy efficient opportunistic routing in underwater WSN B Priyalakshmi, S Murugaveni Wireless Personal Communications 128 (3), 2039-2072 , 2023 2023 Citations: 18
Node replication attack detection in distributed wireless sensor networks L Sujihelen, R Boddu, S Murugaveni, M Arnika, A Haldorai, PCS Reddy, ... Wireless Communications and Mobile Computing 2022 (1), 7252791 , 2022 2022 Citations: 63
Sparse Code Multiple Access for Visible Light Communication and 5G and IoT Application S Murugaveni, B Priyalakshmi, MVS Rahul, S Vasudevan, R Varun Internet of Things and Its Applications, 467-475 , 2021 2021 Citations: 1
RETRACTED ARTICLE: A novel approach for non-orthogonal multiple access for delay sensitive industrial IoT communications for smart autonomous factories S Murugaveni, K Mahalakshmi Journal of Ambient Intelligence and Humanized Computing 12 (6), 6865-6872 , 2021 2021 Citations: 9
WITHDRAWN: Robokart for visually impaired people B Priyalakshmi, S Murugaveni, S Umamaheswari Materials Today: Proceedings , 2021 2021 Citations: 1
Optimal Frequency Reuse Scheme Based On Cuckoo Search Algorithm in Li-Fi 5g Bidirectional Communication K Murugaveni,S, Mahalakshmi IET Communications , 2020 2020 Citations: 13
Survey on efficient use of spatial reusability in multhop wireless network S Murugaveni, K Mahalakshmi Int J Eng Technol 7 (2.24), 431-435 , 2018 2018 Citations: 1
MOST CITED SCHOLAR PUBLICATIONS
Node replication attack detection in distributed wireless sensor networks L Sujihelen, R Boddu, S Murugaveni, M Arnika, A Haldorai, PCS Reddy, ... Wireless Communications and Mobile Computing 2022 (1), 7252791 , 2022 2022 Citations: 63
Emperor penguin optimized Q learning method for energy efficient opportunistic routing in underwater WSN B Priyalakshmi, S Murugaveni Wireless Personal Communications 128 (3), 2039-2072 , 2023 2023 Citations: 18
Optimal Frequency Reuse Scheme Based On Cuckoo Search Algorithm in Li-Fi 5g Bidirectional Communication K Murugaveni,S, Mahalakshmi IET Communications , 2020 2020 Citations: 13
RETRACTED ARTICLE: A novel approach for non-orthogonal multiple access for delay sensitive industrial IoT communications for smart autonomous factories S Murugaveni, K Mahalakshmi Journal of Ambient Intelligence and Humanized Computing 12 (6), 6865-6872 , 2021 2021 Citations: 9
Layering of edge node for jamming attack detection and elimination in wireless sensor network S Murugaveni, B Priyalakshmi Concurrency and Computation: Practice and Experience 35 (22), e7737 , 2023 2023 Citations: 8
Fetal Health Prediction: Comparative Analysis of Machine Learning Models S Sheron, N Sreejith, AB Chandran, S Murugaveni 2025 International Conference on Recent Advances in Electrical, Electronics … , 2025 2025 Citations: 1
RETRACTED ARTICLE: A novel approach of low complexity distributed UA algorithm is used for traffic load balancing and interference in next generation networks E Elamaran, S Murugaveni, S Jyothi, MR Prabhu, MP Chitra, V Talasila Optical and Quantum Electronics 56 (4), 525 , 2024 2024 Citations: 1
Retraction Note to: A novel approach for non-orthogonal multiple access for delay sensitive industrial IoT communications for smart autonomous factories S Murugaveni, K Mahalakshmi Journal of Ambient Intelligence and Humanized Computing 14 (Suppl 1), 151-151 , 2023 2023 Citations: 1
Sparse Code Multiple Access for Visible Light Communication and 5G and IoT Application S Murugaveni, B Priyalakshmi, MVS Rahul, S Vasudevan, R Varun Internet of Things and Its Applications, 467-475 , 2021 2021 Citations: 1
WITHDRAWN: Robokart for visually impaired people B Priyalakshmi, S Murugaveni, S Umamaheswari Materials Today: Proceedings , 2021 2021 Citations: 1
Survey on efficient use of spatial reusability in multhop wireless network S Murugaveni, K Mahalakshmi Int J Eng Technol 7 (2.24), 431-435 , 2018 2018 Citations: 1
Spatio-Temporal Polyp Detection in Colonoscopy Videos Using EfficientNet-B0 and Temporal Transformer A Sahadev, H Tharun, A Tiwari, S Murugaveni 2026 International Conference on Recent Advances in Electrical, Electronics … , 2026 2026
Utilizing Explainable Artificial Intelligence for Parkinson's Disease Diagnosis through the Analysis of Spiral and Wave Drawings with Integrated Data Augmentation S Murugaveni, H Kuresan, NSSC Reddy, P Manoj, G Rohit Advances in Computational Intelligence for Health Informatics and Computer … , 2025 2025
Deep Learning-Based Multiclass Classification for Dental Disease Detection Using DenseNet-201 AB Chandran, S Sheron, N Sreejith, VM Harshini, S Murugaveni International Conference on Data Science and Management, 215-227 , 2024 2024
Real-Time Detection of Anxiety and Panic Attacks ASA Princella, H Kuresan, M Anand, S Murugaveni International Conference on Data Science and Management, 269-281 , 2024 2024
Evaluating the Effectiveness of Deep Learning Models in Parkinson’s Disease Diagnosis Using Magnetic Resonance Imaging S Balamithra, H Kuresan, S Murugaveni, VM Harshini International Conference on Data Science and Management, 241-253 , 2024 2024
Retraction Note: A novel approach of low complexity distributed UA algorithm is used for traffic load balancing and interference in next generation networks E Elamaran, S Murugaveni, S Jyothi, MR Prabhu, MP Chitra, V Talasila Optical and Quantum Electronics 56 (10), 1679 , 2024 2024
