Sworna Kokila M L

Scopus Publications

Enhancing power systems with AI: Design to emission reduction
N. Karpagam, M. L. Sworna Kokila, Bibin Christopher V., Nidhi Kunar, M. D. Mohan Gift, Sampath Boopathi
Innovations in Power Systems and Applications, 2025
Artificial intelligence integration into power systems has been the revolution that transformed how energy is generated, distributed, and consumed. In this regard, this chapter discusses AI-driven methodologies for power system design, optimization, and operation with regards to their potential to reduce carbon emissions. Some of the key applications in this regard include predictive maintenance, smart grid management, and energy demand forecasting, all of which work towards improving system reliability and minimizing waste energy. Advanced AI models, including machine learning and deep learning, allow for real-time decision-making, optimization of renewable energy integration, and dynamic load balancing. They support the installation of distributed energy resources, including solar and wind, which promotes the shift towards cleaner energy systems. The chapter advances how AI can spur transformative reductions in greenhouse gas emissions while paving the way to resilient, intelligent, and sustainable power systems by addressing challenges such as system stability and scalability.
Optimizing Image Retrieval in Cloud Servers with TN-AGW: A Secure and Efficient Approach
N. P. Ponnuviji, G. Nirmala, M. L. Sworna Kokila, S. Indra Priyadharshini
Journal of the Institution of Engineers India Series B, 2025
Quantum-Secured E-Commerce with Blockchain, BB84 and BLAKE3 Encryption using GNN
Harikrishnan S, Gopikrishnan R, M. L. Sworna Kokila, V. Bibin Christopher
Proceedings of 8th International Conference on Inventive Computation Technologies Icict 2025, 2025
The rapid advancements in quantum computing present significant challenges to traditional cryptographic techniques, posing risks to the security of e-commerce transactions. Existing systems are vulnerable to quantumenabled cyberattacks, especially as classical encryption methods, such as RSA and ECC, become obsolete in the face of quantum algorithms. In response, the research proposes an innovative quantum-secure e-commerce fraud detection framework that integrates Graph Neural Networks (GNNs) with Quantum Key Distribution (BB84 QKD), Quantum-Proof-of-Work (QPoW), blockchain technology, and post-quantum cryptographic techniques, including BLAKE3 encryption. The key challenge addressed in the work is ensuring transaction security and fraud detection in a quantum-enabled environment, where classical cryptographic systems are no longer sufficient. Traditional fraud detection mechanisms also struggle to detect subtle anomalies in complex, quantumsecured transaction networks. This approach overcomes these issues by leveraging deep learning-based anomaly detection through GNNs, which model blockchain transactions as graphs to capture intricate relationships between entities. The combination of BB84 QKD ensures secure key distribution, while QPoW offers a resilient consensus mechanism, and BLAKE3 encryption provides fast and secure hashing. The framework aims to offer a robust, scalable solution for ecommerce platforms, financial institutions, and digital identity systems by combining quantum-secure encryption with advanced machine learning for fraud detection. By addressing the growing threat of quantum computing, the methodology provides enhanced security for sensitive transactions, ensuring long-term protection against future quantum-enabled attacks.
MMoE-based Multi-Task Neural Architecture for Cross-Objective Laptop Recommendation
B. Mohana Nagalakshmi, D. Devi Shree, Gifty G, G. Vinoth Rajkumar, S. Sahebzathi, M L Sworna Kokila
Proceedings of the 4th International Conference on Intelligent Computing Information and Control Systems Icoiics 2025, 2025
This research presents an MMoE-based multi-task neural architecture for cross-objective laptop recommendation that simultaneously optimizes user satisfaction across multiple dimensions, including price, performance, portability, and warranty support. The system integrates five complementary neural paradigms: Multi-gate Mixture-of-Experts (MMoE), Deep & Cross Network (DCN), Wide & Deep Network, Neural Collaborative Filtering (NCF), and Graph Neural Network (GNN). Each component captures explicit feature interactions, collaborative signals, relational user-item dependencies, and high-level nonlinear patterns. The MMoE layer enables taskspecific gating of shared experts, allowing the architecture to learn distinct yet interdependent representations for each objective, thereby managing trade-offs between competing factors. Outputs from all models are fused through an adaptive weighting mechanism dynamically adjusted based on contextual embeddings from user behavior and laptop attributes. Representation learning leverages structured specifications such as CPU, GPU, RAM, storage, and battery, along with unstructured inputs like textual reviews and brand feedback, producing highly personalized and interpretable recommendations. Extensive evaluation on a large-scale laptop dataset demonstrates superior performance in ranking metrics, recall, and user satisfaction compared to state-of-the-art baselines. This hybrid architecture offers a scalable, contextaware framework for real-world e-commerce recommendation scenarios.
Deep Learning Algorithms for Object Detection in Smart Environments
D. Dhanya, R. Rajitha Jasmine, M. L. Sworna Kokila, M. Sakthivel, N. Divya, S. Boopathi
Navigating Challenges of Object Detection Through Cognitive Computing, 2025
This chapter has presented the deep learning algorithms in smart environments for object detection purpose; it has, however, come to bode to broader introspect towards more enhanced automation and intelligence. These include advanced architectures of CNN, Region-based CNN, and YOLO which have accomplished their efficiencies in pursuance both for real-time object identification and their subsequent tracking. These offer high accuracy and multiple applications in the smart home, smart city and Industrial IoT, leveraging large datasets and complex computational power. Discussion The discussion addresses challenges in computing complexity, energy efficiency, and model scalability in resource-impoverished environments. Therefore, important case studies and practical examples would describe the ways in which these alleys can be combined with sensor networks and IoT systems, especially the opportunity to revolutionize areas such as security, automation, and adaptable resource management.
Deep Learning-Powered Edge Analytics for IoT-based Sensor Networks
M. Saravana Karthikeyan, Ponnusamy Subramani, M. Panneer Selvam, M. L. Sworna Kokila, R. Vijayalakshmi, P. Sundaravadivel
2nd International Conference on Machine Learning and Autonomous Systems Icmlas 2025 Proceedings, 2025
The rapid expansion of loT-based sensor networks has necessitated the development of efficient edge analytics frameworks to process vast amounts of data in real time while minimizing computational overhead. Deep learning-powered edge computing has emerged as a transformative approach, enhancing data processing capabilities at the network edge while addressing constraints related to latency, bandwidth, and energy consumption. This study introduces a novel algorithm, Lightweight AI Model Compression for Energy-Efficient IoT Edge Devices (LiteEdgeAI), designed to optimize deep learning models for resource-constrained edge environments. The proposed method employs quantization, pruning, and knowledge distillation techniques to achieve significant reductions in model size and computational complexity while maintaining high inference accuracy. To validate the efficiency of LiteEdgeAI, a comparative simulation analysis was conducted against existing state-of-the-art algorithms using key performance metrics such prediction accuracy, energy efficiency and computational complexity. Experimental results demonstrated that LiteEdgeAI outperformed baseline models. The findings highlight the potential of LiteEdgeAI in enabling scalable, low-power, and real-time edge intelligence for IoT-based sensor networks.
Enhanced power system fault detection using quantum-AI and herd immunity quantum-AI fault detection with herd immunity optimisation in power systems
M. L. Sworna Kokila, V. Bibin Christopher, G. Ramya
Iet Quantum Communication, 2024
Quantum computing and deep learning have recently gained popularity across various industries, promising revolutionary advancements. The authors introduce QC‐PCSANN‐CHIO‐FD, a novel approach that enhances fault detection in electrical power systems by combining quantum computing, deep learning, and optimisation algorithms. The network, based on a Pyramidal Convolution Shuffle Attention Neural Network (PCSANN) optimised with the Coronavirus Herd Immunity Optimiser, shows promising results. Initially, historical datasets are used for fault detection. Preprocessing, which includes handling missing data and outliers using Adaptive Variational Bayesian Filtering is followed by Dual‐Domain Feature Extraction to extract grayscale statistical features. These features are processed by PCSANN to detect faults. The Coronavirus Herd Immunity Optimisation Algorithm is proposed to optimise PCSANN for precise fault detection. Performance of the proposed QC‐PCSANN‐CHIO‐FD approach attains 24.11%, 28.56% and 22.73% high specificity, 21.89%, 23.04% and 9.51% lower computation Time, 25.289%, 15.35% and 19.91% higher ROC and 8.65%, 13.8%, and 7.15% higher Accuracy compared with existing methods, such as combining deep learning based on quantum computing for electrical power system malfunction diagnosis (QC‐ANN‐FD), electrical power system fault diagnostics using hybrid quantum‐classical deep learning (QC‐CRBM‐FD), applications of machine learning to the identification of power system faults: Recent developments and future directions (QC‐RF‐FD).
Distributed technologies using ai/ml techniques for healthcare applications
B. Gopi, M. L. Sworna Kokila, Christopher V. Bibin, D. Sasikala, Eric Howard, S. Boopathi
Social Innovations in Education Environment and Healthcare, 2024
The healthcare sector has benefited greatly from the integration of AI/ML with distributed technologies like edge computing, blockchain, and Internet of Things (IoT) to address challenges like data interoperability, security, and scalability. This synergy has a major impact on patient care, medical research, and the efficiency of the healthcare system. AI/ML techniques are used in a variety of fields, including drug development, medical imaging interpretation, picture identification, predictive analytics, and sickness prediction. The relationship between AI/ML and distributed technologies—such as decentralized architectures for safe access to real-time data sources, blockchain for data integrity and privacy, and edge computing for low-latency processing—is discussed. When combining AI/ML with dispersed technology, the healthcare business faces trends and concerns related to interoperability, legal compliance, and ethical issues.
Securing cloud-based medical data: an optimal dual kernal support vector approach for enhanced EHR management
M. L. Sworna Kokila, E. Fenil, N. P. Ponnuviji, G. Nirmala
International Journal of System Assurance Engineering and Management, 2024
Optimizing Energy Consumption in Smart Grids Using Demand Response Techniques
SwornaKokila M L, Venkatarathinam R, Rose Bindu Joseph P, M. A. Manivasagam, Kakarla Hari Kishore
Distributed Generation and Alternative Energy Journal, 2024
Smart grids have developed as a potentially game-changing strategy for controlling the demand and supply of energy. Unfortunately, peak demand is a significant source of grid instability and rising energy prices, making it one of the most critical difficulties in smart grids. During times of high energy demand on the grid, demand response (DR) strategies incentivize consumers to change how they use energy. This study’s overarching goal is to learn how DR methods may be used to help smart grids make better use of their energy resources. The primary research is to develop a smart DR system that can predict times of high energy demand and proactively alter usage to reduce such periods. Machine learning strategies are utilized in the proposed system to estimate peak demand via past data, weather predictions, and other variables. The system will then alter energy use based on real-time data from smart meters along with other sensing devices to meet the projected demand. The simulation model will include several scenarios for testing the DR system’s flexibility, including a range of weather conditions, load profiles, and grid topologies. Several indicators, including peak demand reduction (80.04%), energy savings (38.09%), environmental consequences, and reaction time (<0.4 seconds), are used to evaluate the model’s performance. The output of the method excelled all of the other current methods that were taken into account. The system’s rapid response time and its positive environmental impact further highlight its potential in managing smart grid resources effectively.
Blockchain-Enabled Energy Trading in Microgrids for Sustainable Computing Infrastructure
M. L. Sworna Kokila, Sunitha R, P. Kalyanakumar, J. Relin Francis Raj, R. Santhana Krishnan, S. Gopikumar
5th International Conference on Electronics and Sustainable Communication Systems Icesc 2024 Proceedings, 2024
Power-Optimized Process Management in Heterogeneous Computing Environments
R. Santhana Krishnan, G. Vinoth Rajkumar, M L Sworna Kokila, J. Relin Francis Raj, S. Jegadeesan, S. Gopikumar
2nd International Conference on Sustainable Computing and Smart Systems Icscss 2024 Proceedings, 2024
Efficient abnormality detection using patch-based 3D convolution with recurrent model
M. L. Sworna Kokila, V. Bibin Christopher, R. Isaac Sajan, T. S. Akhila, M. Joselin Kavitha
Machine Vision and Applications, 2023
A unique approach of person reidentification using auto track regression framework
M.L. Sworna Kokila, V. Gomathi
Journal of Intelligent and Fuzzy Systems, 2022

Sworna Kokila M L

Scopus Publications

RECENT SCHOLAR PUBLICATIONS

MOST CITED SCHOLAR PUBLICATIONS