Improving routing in wireless sensor networks technology on data aggregation using fused cluster routing algorithm G. S. Pradeep Ghantasala, L. Godlin Atlas, R. Sathiyaraj, S. Ponmaniraj, Pellakuri Vidyullatha, M. Arvindhan Discover Sustainability, 2025 A crucial challenge in wireless sensors networks (WSNs) is effective communication with tolerable power depreciation of sensor nodes. A bottleneck in the transmission of superfluous data is inevitable in every large network. Energy, latency, and data redundancy are the three identified challenges; to increase existing effectiveness, a trade-off must be made. Existing methods in the literature either employ a distributed or centralized strategy to select a sensor node. The paper proposes a fused clustering-based model with a routing algorithm. In this paper, a dedicated energy-efficient Fused Cluster Routing Algorithm (FCRA) intends to improve data aggregation in WSNs. With applications such as environmental monitoring, catastrophe management, smart agriculture, and urban planning being deployed, a need for low-energy communication protocols breeds more essential. The proposed FCRA integrates hierarchical clustering to optimize routing, thereby reducing transmit redundancy and conserving node energy. This directly contributes to the longevity and reliability of sensor networks in the domains of sustainability. The ideal routing technique for Cluster-Head (CH) identification is dependent on every sensor node’s processing capability, and it also assists in handling the intermediaries in the communication path. To enable efficient communication, data from both clusters is gathered utilizing location and temporal correlation, then aggregated. While related to the earlier suggested methods, the proposed method efficacy is shown to be higher, indicating its effectiveness. The efficacy of the proposed algorithm is assessed using modelling criteria such as aggregation, packet transfer percentage, throughput, packet delay, and used energy. This tremendous finding lays emphasis on optimized routing of WSNs which aids in sustainable infrastructures, resilient ecosystems, and informed environmental governance.
Multifactor Authentication (MFA), the golden lock for cloud entry: By adopting MFA, organizations and individuals can significantly reduce the risk of security breach Arvindhan Muthusamy, G. Sakthi Risk Based Approach to Secure Cloud Migration, 2025 Today's increasingly frequent and costly cyber threats underscore the importance of securing machine learning (ML), which can automate reactions to identified risks and decrease the time needed to counteract possible assaults. Steps like separating infected systems and launching predetermined security procedures fall under this category. Machine learning algorithms sift through mountains of data, searching for irregularities that indicate potential security risks. Unusual user activity, malware, and phishing attempt detection are all part of this. To verify identities, ML models examine patterns of user behaviour. Some examples of this behaviour are the dynamics of keystrokes and mouse movements. In every respect, current MFA solutions are more user- and admin-friendly. More alternatives mean businesses can pick an MFA solution that suits their specific setup and requirements. A user's knowledge, possession, or status can be validated via MFA methods.
Cybersecurity for preserving mental wellness and preventing abuse Arvindhan Muthusamy, Tarun Kumar, Minakshi Exploiting Machine Learning for Robust Security, 2025 The rapid surge in cybercrime during the COVID-19 pandemic has become a pressing concern for mental health professionals. The increased reliance on digital technologies for work, education, and healthcare has led to a rise in cyber-attacks, with perpetrators exploiting vulnerable individuals through methods like phishing scams and online harassment. These incidents can have severe psychological consequences, such as heightened anxiety, depression, and post-traumatic stress. To address this challenge, the text explores the dynamic interplay between cybersecurity practices and the preservation of mental health and well-being. It examines various cyber threats, including cyberbullying, doxing, and revenge porn, and their emotional and mental impacts on victims, such as anxiety, depression, and social withdrawal.
Smart Farming Through Multi-Modal Attention Networks: Climate and Yield Forecasting from IoT Sensors Thrilok Kolla, GS Pradeep Ghantasala, Shaik Sharmila, Konda Manasa, Anuradha Reddy, M. Arvindhan Proceedings of International Conference on Digital Innovations for Sustainable Solutions Icdiss 2025, 2025 Integration of emerging technologies, such as the Internet of Things in sensing and machine learning in analyzing remote-sensing photographs, is opening new avenues for precision agriculture in the management of field crop production in an especially timely and data-driven manner. Unfortunately, agricultural datasets are heterogeneous and high-dimensional, including static soil properties, dynamic meteorological sequences, and multi-resolution images. Hence, none of these methods has been developed so far to understand the hierarchical time-space dependencies of such data, apart from creating interpretable outputs for decision-makers. With this, we have proposed the development of a Hierarchical Attention Network for Climate and Yield Impact Prediction that is, in itself, a promising novel articulated end-to-end deep learning architecture for universality in realtime agriculture predictability in multi-modal aspects. It employs a multi-level attention mechanism to capture temporal evolutions in each modality and combines ensemble-inspired diversity for further robustness. The tendency is that attention becomes part of a prediction process, under the rationale of primary drivers of yield variability, thereby creating transparency for decisionmaking. Empirical validation through widely diverse agricultural datasets showed that the new model significantly enhanced predictor performance and interpretability vis-à-vis cutting-edge techniques in this field, and as such, best fit in operational deployment under those climates for accuracy improvements in precision agriculture resilience.
Optimization of Food Preservation Techniques Using Multi-Objective Evolutionary Algorithms and Soft Computing Models Thrilok Kolla, G S Pradeep Ghantasala, A. Ravi Kumar, A. V. Lakshmi Prasuna, Anuradha Reddy, M. Arvindhan Proceedings of International Conference on Digital Innovations for Sustainable Solutions Icdiss 2025, 2025 Identifying water pollution in complex river systems poses significant challenges due to sparse monitoring infrastructure, heterogeneous data sources, and measurement uncertainties. Traditional deterministic approaches struggle to provide reliable source attribution when faced with incomplete datasets and dynamic spatiotemporal dependencies. This paper presents a Bayesian network-assisted framework for pollutant source identification that integrates multi-source data streams, including insitu water quality measurements, hydrological simulations, and remote sensing observations into a unified probabilistic model. The proposed methodology explicitly incorporates river network topology and temporal evolution dynamics while handling uncertainty through principled Bayesian inference. The framework employs a linear advection-decay model to simulate pollutant transport, combined with maximum a posteriori estimation to identify the most probable source configurations. Feature selection incorporates temporal derivatives, upstream influence metrics, and lagged cross-correlations to capture spatiotemporal dependencies effectively. Experimental validation on synthetic river networks demonstrates the framework's capability to identify pollution sources even under noisy measurement conditions accurately. The probabilistic approach provides interpretable uncertainty quantification, enabling robust decision support for water quality managers. Results show superior performance compared to traditional deterministic methods, particularly in scenarios with limited sensor coverage. The framework offers a scalable, interpretable solution for environmental monitoring and pollution control, providing actionable insights for timely remediation strategies and evidence-based policy formulation in water resource management.
Transformer-Augmented Spatio-Temporal Deep Neural Architectures for Multimodal Monitoring and Predictive Analytics in Fruit Cultivation Systems Thrilok Kolla, G S Pradeep Ghantasala, Aswathi S. Sheelan, Prashanth Kumar P, Anuradha Reddy, M. Arvindhan Proceedings of International Conference on Digital Innovations for Sustainable Solutions Icdiss 2025, 2025 Accurate prediction of flowering times in ornamental species is essential for ecological research, cultural planning, and climate change adaptation. Conventional phenological models often rely on single data sources such as temperature or precipitation, limiting their predictive power in complex ecosystems. The proposed study introduces a novel multimodal big-data fusion framework that integrates climatic indicators (Growing Degree Days, chill units, temperature, precipitation), geographic variables (latitude, elevation, heat zones), vegetation indices (NDVI/EVI), and species-specific encodings for high-resolution forecasting of flowering day-of-year (DOY). By combining diverse data modalities, the model captures cross-dimensional interactions that more closely represent the biological and environmental drivers of phenology. The framework employs interpretable machine learning, rigorous preprocessing, temporal alignment of vegetation indices, and feature selection to build robust predictive models. Results demonstrate superior performance with an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{R}^{\mathbf{2}}$</tex> of 0.92, RMSE of 4.3 days, and MAE of 3.1 days across multiple species. Growing Degree Days emerged as the strongest predictor, while NDVI signals enhanced fine-scale temporal alignment. Error distributions concentrated near zero further validated the framework's reliability and scalability. This research presents an approach to creating transferable and interpretable explanations of phenological models. Its framework enables accurate predictions of flowering events, thereby contributing to the conservation of biodiversity, development, and landscaping in urban environments, as well as planning for climate-resilient agriculture for the greater benefit of society.
Spam Detection Using Machine Learning Anukool, Vijendra Vikram Chauhan, Arvindhan M, Rajat Subhra Chakraborty, G. Sudeshna Chakraborty Proceedings IEEE 2024 1st International Conference on Advances in Computing Communication and Networking Icac2n 2024, 2024
Human Routine Analyzer Using Machine Learning M. Arvindhan, Akshat Sharma, Atul Kumar Prasad Proceedings IEEE 2023 5th International Conference on Advances in Computing Communication Control and Networking Icac3n 2023, 2023
Breast Cancer Detection using Machine Learning Bipul Raj, Kishlay Raj, L. Vetrivendan, M. Arvindhan Proceedings IEEE 2023 5th International Conference on Advances in Computing Communication Control and Networking Icac3n 2023, 2023
