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


Palanivel Sivakumar has a total of 144 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Network Security and Privacy Protection in Cyberattacks With Asynchronous Reinforcement Federated Learning With Task Offloading: Decentralized Real‐Time Iteration Approach

    by S. Nandhini, Sivakumar Poruran, Somasundaram Devarajan

    Medical healthcare has experienced a revolution through the Internet of Medical Things (IoMT) system, which needs crucial improvements for security measures while working towards better privacy and efficiency capabilities. This paper develops asynchronous reinforcement federated learning (FL) with task offloading (ARFL‐TO) as a combination of FL with reinforcement learning (RL) through dynamic TO systems to boost scalability alongside adaptability and security in heterogeneous environments, including autonomous systems, smart grids, and industrial Internet of Things (IIoT) settings. ARFL‐TO requires validation through different datasets, which encompass a range of disease groups together with multiple medical instruments. The analysis indicates that ARFL‐TO surpasses FL‐TO and federated reinforcement‐based fusion (FRF) in productivity by 42.22% and reduces power requirements by 79.22% while shortening processing time by 7.13% during real‐world operation, where networks become unstable and data transmission is affected. This framework achieves privacy protection through secure data pooling, and it makes models understandable for clinical support in addition to delivering enhanced energy efficiency for low‐power medical devices. Future investigation directs toward integrating the proposed system into real healthcare environments and developing performance improvements for severe operational scenarios, and introducing adaptable hyperparameter techniques suitable for healthcare systems requiring dynamic adjustments. ARFL‐TO establishes itself as an efficient framework for secure time‐sensitive decision‐making across decentralized medical networks and cross‐sectoral systems, maintaining privacy concerns.


  • Open Access

    Enhancing Intrusion Detection with Dimensionality Reduction Methods Using Machine Learning

    by M M.Rathika, P P.Sivakumar, V V.Bhuvaneshwari, S. Sandosh

    This project investigates the efficacy of machine learning algorithms, including Naive Bayes (NB), NLP, and K-Nearest Neighbors (KNN), in the context of Intrusion Detection Systems (IDS), with a focus on dimensionality reduction techniques. By leveraging datasets such as USNW NB 15, NSL-KDD, and UNSWNB15, the research aims to assess the accuracy and efficiency of these algorithms in identifying malicious network activities while integrating Dimensionality reduction methods. Through systematic experimentation and evaluation, the project seeks to determine the most effective algorithm for intrusion detection, considering both traditional metrics and the impact of Dimensionality reduction. The findings contribute to enhancing internet security measures by identifying robust solutions for detecting and preventing intrusions amidst complex network traffic. This research underscores the importance of incorporating dimensionality reduction techniques to improve the performance of IDS and mitigate potential threats effectively. The project provides insights for practitioners and researchers in selecting appropriate algorithms and integrating dimensionality reduction strategies for more efficient intrusion detection systems.

  • Open Access

    Enhanced Disease Identification Model for Tea Plant Using Deep Learning

    by Santhana Krishnan Jayapal, Sivakumar Poruran

    Tea plant cultivation plays a significant role in the Indian economy. The Tea board of India supports tea farmers to increase tea production by preventing various diseases in Tea Plant. Various climatic factors and other parameters cause these diseases. In this paper, the image retrieval model is developed to identify whether the given input tea leaf image has a disease or is healthy. Automation in image retrieval is a hot topic in the industry as it doesn’t require any form of metadata related to the images for storing or retrieval. Deep Hashing with Integrated Autoencoders is our proposed method for image retrieval in Tea Leaf images. It is an efficient and flexible way of retrieving Tea Leaf images. It has an integrated autoencoder which makes it better than the state-of-the-art methods giving better results for the MAP (mean average precision) scores, which is used as a parameter to judge the efficiency of the model. The autoencoders used with skip connections increase the weightage of the prominent features present in the previous tensor. This constitutes a hybrid model for hashing and retrieving images from a tea leaf data set. The proposed model will examine the input tea leaf image and identify the type of tea leaf disease. The relevant image will be retrieved based on the resulting type of disease. This model is only trained on scarce data as a real-life scenario, making it practical for many applications.


  • Open Access

    Cooperative Communications Based on Deep Learning Using a Recurrent Neural Network in Wireless Communication Networks

    by M. Rathika, P. Sivakumar, K. Ramash Kumar, Ilhan Garip

    In recent years, cooperative communication (CC) technology has emerged as a hotspot for testing wireless communication networks (WCNs), and it will play an important role in the spectrum utilization of future wireless communication systems. Instead of running node transmissions at full capacity, this design will distribute available paths across multiple relay nodes to increase the overall throughput. The modeling WCNs coordination processes, as a recurrent mechanism and recommending a deep learning-based transfer choice, propose a recurrent neural network (RNN) process-based relay selection in this research article. This network is trained according to the joint receiver and transmitter outage likelihood and shared knowledge, and without the use of a model or prior data, the best relay is picked from a set of relay nodes. In this study, we make use of the RNN to do superdimensional (high-layered) processing and increase the rate of learning and also have a neural network (NN) selection testing to study the communication device, find out whether or not it can be used, find out how much the system is capable of, and look at how much energy the network needs. In these simulations, it has been shown that the RNN scheme is more effective on these targets and allows the design to keep converged over a longer period of time. We will compare the accuracy and efficiency of our RNN processed-based relay selection methods with long short-term memory (LSTM), gated recurrent units (GRU), and bidirectional long short-term memory (BLSTM),which are all acronyms for long short-term memory methods.


  • Smart bin: Waste segregation system using deep learning‐Internet of Things for sustainable smart cities

    by K.O. Mohammed Aarif, Caffiyar Mohamed Yousuff, B.A. Mohammed Hashim, C. Mohamed Hashim, P. Sivakumar

    Waste management is a major issue with the emerging growth in the world population, and we need to find efficient ways to recycle and reuse waste. Segregating waste has become a primary need in waste management as different types of waste like Bio & Non‐Bio‐degradable waste should be processed differently. Effective waste isolation at the fundamental level is especially required for this. Several Smart cities oriented smart garbage management systems are also proposed using Internet of Things (IoT) and GSM. The existing smart bins using IoT and wireless sensor network (WSN) are dependent significantly on two major things. First, multiple types of sensors, as a single sensor may not be able to detect different material waste, and second, the console (Microcontroller, Arduino Raspberry Pi) and connectivity which in turn dependent on programming and operating system. These limitations of the embedded smart bin are overcome by combining IoT with artificial intelligence approaches such as deep neural network (DNN) systems. In this paper, we have presented a Friendly Waste Segregator Using Deep Learning and the IoT to classify and isolate the waste objects as biodegradable and nonbiodegradable. Our proposed method utilizes, a robust deep learning network to classify the waste accurately and IoT for monitoring and connectivity using various sensors. Our proposed method with initial training can identify and segregte real‐time waste objects without human intervention with an average accuracy of 97.49 %. Our smart bin intends to provide optimized waste management of bio and non‐bio‐waste and help to build an ecologically safe society.


  • An efficient approach for removing haze from single image using Gaussian pyramidal decomposition

    by Senthilkumar Kuppusamy Pandian, Sivakumar Poruran

    Summary: The visibility of the image is degraded by haze components which are affected in image during image acquisition process. The haze removal process can be categorized into haze reduction from single image and haze reduction from multiple images. The second case consumes more time and the complexity of algorithm is high. In order to overcome such limitations, this paper proposes an efficient methodology for haze reduction in single acquired image. In this paper, haze detection and removing methodology is proposed using Gaussian pyramidal decomposition. This proposed methodology is tested on both indoor and outdoor haze affected images. The performance of this proposed haze removal algorithm is analyzed with respect to average running time and peak signal to noise ratio, mean square error, mean absolute error, Entropy, and normalized histogram intersection coefficient.


  • Detection and Segmentation of Cracks in Weld Images Using ANFIS Classification Method

    by L. Mohana Sundari, P. Sivakumar

    This paper proposes the detection and classifications of weld images for crack detection using image processing techniques. The proposed method consists of preprocessing stage, feature extraction stage, classification stage and crack region segmentation regions. The image enhancement method is used as preprocessing stage and texture and statistical features are extracted from the enhanced weld images. These computed features are then classified into “Excess weld”, “Good weld”, “No weld” and “Undercut weld”, using Adaptive Neuro Fuzzy Inference System (ANFIS) classification method. This proposed method is analyzed in terms of sensitivity, specificity, accuracy, positive predictive value, negative predictive value and precision. The simulation results of the proposed method are compared with other state of the art methods.


  • A Novel Deep Convolutional Neural Network Architecture Based on Transfer Learning for Handwritten Urdu Character Recognition

    by Mohammed Aarif Kilvisharam Oziuddeen, Sivakumar Poruran, Mohamed Yousuff Caffiyar

    Deep convolutional neural networks (CNN) have made a huge impact on computer vision and set the state-of-the-art in providing extremely definite classification results. For character recognition, where the training images are usually...