Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Krishnamoorthy Parkavi has a total of 25 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Software-defined Networking Controller for Detection of DDoS Attacks Based on Deep Neural Networks

    by Jeyavim Sherin R. C., Parkavi K.

    Advancements in technology contribute to an increased vulnerability to cyberattacks, with Distributed Denial of Service (DDoS) attacks being a prominent threat. Attackers overwhelm network servers with excessive data, hindering legitimate users from accessing them. Software Defined Networking (SDN) is particularly susceptible due to its centralized architecture, making it a prime target for DDoS attacks aimed at the control planes. As cloud computing has grown rapidly, software-defined networks have been developed to provide dynamic management and enhanced performance. Several security concerns are growing, especially as DDoS attacks and malicious actors become more interested in SDN controllers. Many researchers have proposed detecting DDoS attacks. Due to their unqualified features and non-realistic data sets, these approaches have high false positive rates and low accuracy. As a result, SDN controllers can be protected against DDoS attacks using deep learning algorithms (DL). Furthermore, the suggested method comprises three phases: The process involves pre-processing the data, selecting significant features for DDoS detection based on correlation, and utilizing Deep Neural Networks (DNNs) for the detection. In order to evaluate the efficiency of the method proposed, we employ a benchmarking dataset to evaluate the false positive rate as well as detectability, with the traditional assessment indicators. In this paper, we propose a deep learning method for detection of DDoS attacks called DNNADSC, which is the first anomaly detection method based on deep neural network for DDoS attacks. The method proposed efficaciously recognizes DDoS attacks, with the detection rate of 99.39%, with a precision of 97.41% with a false-positive rate (FPR) that is 0.0665 with the F1 measure of 99.32%.


  • Open Access

    Decoy Traffic-Aware IDS: A Novel Approach Using CSHHHNN and GGGOA

    by R. C. Jeyavim Sherin, K. Parkavi

    In the digital world, the Network Intrusion Detection System (NIDS) plays a prime role in protecting against cyber threats. Nevertheless, none of the prevailing studies focused on analyzing the impact of decoy traffic in networks, thereby resulting in poor security. To address this research gap, this article proposes a decoy traffic-aware NIDS that incorporates enhanced Cognitive Spiking Hyperbolic Hodgkin–Huxley Neural Networks (CSHHHNN) and the Greylag Griewank Goose Optimization Algorithm (GGGOA). The process starts with node initialization and registration, followed by network segmentation via clustering and subnet masking schemes. Thereafter, possible routes are recognized for each node, and then, a graph structure is generated. Next, multipath routing is established between the source and destination. Afterward, the network paths are verified. If a valid path is found, then it is selected; otherwise, it is discarded. Now, data sensing and data encryption are done. Thereafter, load balancing is performed in the encrypted data. Then, the load-balanced data is fed into the intrusion detection system. In the proposed NIDS, the dataset is initially pre-processed. Afterward, the relevant features are extracted. Subsequently, optimal features are selected, and privacy measures are applied to the model. Finally, the improved CSHHHNN model is used to classify the network traffic into normal or attack categories. The proposed approach demonstrates superior performance with 99.5124% accuracy in decoy traffic management and intrusion detection.

  • Open Access

    Enhancing Road Safety: Detection of Animals on Highways During Night

    by K. Parkavi, Agniv Ganguly, Shivansh Sharma, Ashesh Banerjee, Keshav Kejriwal

    Animals on highways at night can create serious accidents that put both human and wildlife lives in danger. Conventional monitoring systems are rendered useless due to poor visibility and inadequate lighting, emphasizing the necessity for a sophisticated detecting system. Using YOLOv5, a cutting- edge object detection algorithm, this study presents a way to identify animals on highways at night and reduce collisions. To improve image quality in low-light circumstances, the model uses pre-processing techniques such as Contrast Limited Adaptive Histogram Equalization (CLAHE) and a robust Retinex model. This enhancement enables YOLOv5 to achieve improved detection accuracy on enhanced photos, increasing its effectiveness in spotting animals in low-light situations. YOLOv5 offers quick, highly accurate detection in contrast to conventional techniques like infrared and thermal imaging, which frequently lack the resolution for accurate species identification. The model, trained on a broad dataset, performed well, with a precision of 0.923, recall of 0.773, mean average precision (mAP50) of 0.802, and mean average precision (mAP50-95) of 0.567, demonstrating its competence in spotting animals in normal nighttime circumstances. Despite the model’s good performance in normal low-light situations, problems still arise in unfavorable conditions including rain, fog, and dense foliage, which can lower detection accuracy. This study aims to improve road safety and aid to wildlife conservation by minimizing highway collisions. Subsequent endeavors will center on enhancing the model’s resilience to withstand harsh circumstances and further diminish the likelihood of accidents.


  • Open Access

    An innovative model for an enhanced dual intrusion detection system using LZ‐JC‐DBSCAN, EPRC‐RPOA and EG‐GELU‐GRU

    by Jeyavim Sherin R. C., Parkavi K.

    The rise of suspicious activities in network communication, driven by increased internet accessibility, necessitates the development of advanced intrusion detection systems (IDS). Existing IDS solutions often exhibit poor performance in detecting suspicious activity and fail to identify various attack types within packet capture (PCAP) files, which monitor network traffic. This paper proposes a deep learning‐based dual IDS model designed to address these issues. The process begins with utilizing the CSE‐CIC‐IDS2019 dataset to extract features from PCAP files. Suspicious activities are detected using the Exponential Geometric‐Gaussian Error Linear Units‐Gated Recurrent Unit (EG‐GELU‐GRU) method. Normal data undergoes further feature extraction and preprocessing through Log ZScore‐Jacosine Density‐Based Spatial Clustering of Applications with Noise (LZ‐JC‐DBSCAN). Feature selection is optimized using the Entropy Pearson R Correlation‐Red Panda optimization algorithm. Suspicious files are flagged, while load balancing is performed on normal data. Attack detection is achieved through word embedding with the Glorot Kaufman‐bidirectional encoder representations from transformers technique and classification via the EG‐GELU‐GRU model. Attacked packets are blocked, and the method is reapplied for attack‐type classification. Experimental results using Python demonstrate the model’s superior performance, achieving 98.18% accuracy and 98.73% precision, surpassing existing approaches and significantly enhancing intrusion detection capabilities.


  • IoT Next Gen and the World of Big Data Computing

    by Nirmala Venkatachalam, Leninisha Shanmugam, K Parkavi, D Satheeswari

    Big data and the Internet of Things (IoT) are hastily growing and shaking many elements of technology and industry, enhancing the assistance for both businesses and individual people. Throughout the last few years, as the miniaturisation of IoT devices has increased, massive volumes of data have been created. With the explosive growth of the IoT, Big Data technologies are known as an essential data analytics tool for delivering information into IoT foundations to serve the objective of the IoT approaches and help in key decision making. Despite the fact that big data analytics is a well-studied topic, the diversity of IoT realms has divided the progress of big data approaches in each IoT domain. Mutual understanding amongst IoT disciplines might therefore promote the extension of big data research in IoT. This study analyses state-of-the-art research efforts aimed on big data technologies in various IoT areas to promote and encourage knowledge transfer across IoT domains. Furthermore, discussion of many unique prospects in the IoT area is provided because of data analytics. Lastly, as future study directions, outstanding research problems are recognised and discussed.


  • Methodology for Programming of AI-Based IDS

    by Krishnamoorthy Parkavi, Rajayan Christy Jeyavim Sherin

    In the dynamic world, NIDS (Network Intrusion Detection System) is one of the important methods in cyber security. The intrusion detection system (IDS) is an important “security technology.” The modern techniques – Artificial-Intelligence-based technique, machine learning algorithms, and deep learning approaches – use an efficient selection of features from datasets like KDD CUP99. The deep learning algorithm is a trending technology for securing the network from attacks and enhancing performance. In this chapter, we discuss the main concern in the convergence and speed of IDS. This chapter helps various advanced methodologies in AI with programming an AI-based IDS.

  • Top Cited Papers
    Open Access

    An Artificial-Intelligence-Based Renewable Energy Prediction Program for Demand-Side Management in Smart Grids

    by Vinothini Arumugham, Hayder M. A. Ghanimi, Denis A. Pustokhin, Irina V. Pustokhina, Vidya Sagar Ponnam, Meshal Alharbi, Parkavi Krishnamoorthy, Sudhakar Sengan

    Technology advancements have enabled the capture of Renewable Energy Sources (RES) on a massive scale. Smart Grids (SGs) that combine conventional and RES are predicted as a sustainable method of power generation. Moreover, environmental conditions impact all RES, causing changes in the amount of electricity produced by these sources. Furthermore, availability is dependent on daily or annual cycles. Although smart meters allow real-time demand prediction, precise models that predict the electricity produced by RES are also required. Prediction Models (PMs) accurately guarantee grid stability, efficient scheduling, and energy management. For example, the SG must be smoothly transformed into the conventional energy source for that time and guarantee that the electricity generated meets the predicted demand if the model predicts a period of Renewable Energy (RE) loss. The literature also suggests scheduling methods for demand-supply matching and different learning-based PMs for sources of RE using open data sources. This paper developed a model that accurately replicates a microgrid, predicts demand and supply, seamlessly schedules power delivery to meet demand, and gives actionable insights into the SG system’s operation. Furthermore, this work develops the Demand Response Program (DRP) using improved incentive-based payment as cost suggestion packages. The test results are valued in different cases for optimizing operating costs through the multi-objective ant colony optimization algorithm (MOACO) with and without the input of the DRP.