Papers by Priyadarshini Radhakrishnan
International Journal of HRM and Organizational Behavior, 2023
As the number of Internet of Things (IoT) devices skyrockets, IoT networks are now more susceptib... more As the number of Internet of Things (IoT) devices skyrockets, IoT networks are now more susceptible to cyber-attacks, especially ICMP Flood attacks, which flood the network with too many ICMP Echo Request packets and thus overwhelm it. This research seeks to improve the detection and mitigation of ICMP Flood attacks in real time with minimal computational overhead and fewer false positives. The solution utilizes a hybrid ensemble learning method coupled with a Support Vector Machine (SVM) to efficiently classify network traffic. The system combines several classifiers, such as Decision Tree, Random Forest, and k-Nearest Neighbors (k-NN), to enhance detection accuracy and reliability. Through the use of feature extraction methods, the system differentiates between normal and attack traffic. Experimental findings indicate that the model has 95% accuracy, 94% precision, 96% recall, and 95% F1 score with a low 3% false positive rate and a minimal computational overhead of only 20 milliseconds, making it real-time deployable. The ensemble hybrid model surpasses single classifiers, indicating how effective the use of multiple machine learning methods is for enhanced detection performance. This work shows the capabilities of ensemble learning and SVM for enhancing the security of IoT networks, presenting an efficient and scalable method for the detection and alleviation of ICMP Flood attacks at little computational cost but with high accuracy.
International Journal of Information Technology & Computer Engineering, 2024
In an effort to improve operational effectiveness and strategic decision-making, this study explo... more In an effort to improve operational effectiveness and strategic decision-making, this study explores how Big Data Analytics (BDA) and the Internet of Things (IoT) can be integrated inside the Business Intelligence (BI) framework. In order to handle the enormous datasets produced by IoT devices, the research investigates cutting-edge analytical methods including machine learning and predictive analytics. The suggested framework is evaluated using eight critical performance measures, such as data processing speed, integration efficiency, prediction accuracy, and accuracy of hypothesis testing, versus more established analytical techniques (SmartPLS, SPSS, and PLS-SEM). The IoT-BDA integrated BI framework performs noticeably better than conventional approaches, according to the results, especially in terms of real-time processing and system scalability. Enabling IoT and BDA in BI can result in more accurate, efficient, and scalable data-driven decision-making systems, giving businesses a competitive edge. An ablation study confirms the significance of each component within the
International journal of modern electronics and communication engineering, 2022
The rapid growth of healthcare data and the increasing need for efficient management have led to ... more The rapid growth of healthcare data and the increasing need for efficient management have led to challenges in cloud-based healthcare systems, including scalability, data security, and integration. Existing systems often struggle to manage large data volumes while ensuring secure transmission and storage. The aim of this work is to develop a secure and scalable cloud-based framework for efficient healthcare data collection and monitoring. The framework begins with collecting healthcare data from various sources, followed by preprocessing steps such as K-Nearest Neighbors (k-NN) imputation to handle missing values and Min-Max scaling for normalization. The data is then encrypted using Salsa 20 to ensure security, and Transport Layer Security is applied for secure data transmission to the cloud. The processed data is stored in cloud-based solutions for efficient management and real-time access. The results show that the latency of cloud systems increases with system load, from 1000 ms at lower loads to 4500 ms at higher loads, demonstrating the challenge of maintaining low latency as demand grows. Additionally, the Salsa 20 encryption achieves near 100% security strength with key sizes of 1024 bits. The contribution of this work lies in developing a robust, efficient, and secure framework that enhances healthcare data management while ensuring both performance and data security.
INTERNATIONAL JOURNAL OF APPLIED SCIENCE ENGINEERING AND MANAGEMENT, 2018
The rapid expansion of e-commerce has increased the need for intelligent personalization and dyna... more The rapid expansion of e-commerce has increased the need for intelligent personalization and dynamic pricing strategies to enhance customer engagement and revenue optimization. Traditional recommendation methods rely on collaborative filtering and deep learning models, which often suffer from scalability issues and high computational costs. Similarly, conventional pricing approaches lack adaptability, leading to suboptimal revenue generation. To address these limitations, we propose an AI-powered cloud commerce framework that integrates LightGCN for efficient product recommendations and a Multi-Armed Bandit (MAB) with Thompson Sampling for adaptive dynamic pricing. The approach is lightweight, scalable, and cloud-optimized, reducing computational overhead while maintaining high accuracy. Experimental results demonstrate that the model achieves HR@10 of 85% and NDCG@10 of 78%, significantly outperforming conventional recommendation techniques. Furthermore, the pricing model reduces regret by 25%, optimizing revenue adaptation. Computational efficiency is improved, with 40% fewer FLOPs and 30% lower latency, making it suitable for large-scale applications. Additionally, the cloud-optimized storage strategy results in 70% storage reduction and 60% faster retrieval, enhancing data accessibility. Compared to traditional frameworks, the method delivers higher accuracy, faster decision-making, and superior cloud efficiency, ensuring a competitive edge in AI-driven e-commerce. This work advances the field by bridging recommendation and pricing optimization with lightweight AI, offering an efficient, scalable, and adaptable solution for next-generation e-commerce platforms.
Uploads
Papers by Priyadarshini Radhakrishnan