P. Parthasarathi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
P. Parthasarathi has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- G. Prabu Kanna
School of Computing Science Engineering and Artificial Intelligence (SCAI), VIT Bhopal University, Bhopal, India
2 shared publications - S J K Jagadeesh Kumar
Department of Computer Science and Engineering, Kathir College Of Engineering, Coimbatore, India
2 shared publications S. Nivedha
2 shared publications- Yogesh Kumar
Department of CSE, School of Technology, Pandit Deendayal Energy University, Gandhinagar, India
2 shared publications - .H. Summia Parveen1 shared publication
- G. Sathish Kumar
Department of Artificial Intelligence and Data Science, Sri Eshwar College of Engineering, Coimbatore, India
1 shared publication
Latest Publications
- 30 October 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Software Engineering and Knowledge Engineering
- p. 1-16
An Energy-Efficient Cooperative DDoS Attack Mitigation Using a Federated Deep Learning Framework in SDN
A distributed denial-of-service (DDoS) attack can be launched through the huge number of edge-connected devices on software-defined AIoT systems. The lack of modern labeled training data makes it impossible to keep defending it in a centralized way, and privacy issues prevent the sharing of sensitive traffic information. In this work, a trustworthy federated learning (FL) architecture is proposed to tackle these challenges in a decentralized defense framework for AIoT. The framework incorporates three elements. First, a private model update exchange scheme based on secure multiparty computation can be used to enable privacy-preserving exchange of local model updates, while being robust to outliers. Second, a horizontal cooperation-based mitigation mechanism minimizes packet loss and decreases the effect of the attack on regular traffic. Third, a semi-supervised FL-based anomaly detection strategy is used to correctly classify attack classes and minimize the number of false alarms, based on a limited number of labeled outliers. Experimental evaluation on real-world datasets (CICIDS2017 and InSDN) demonstrates the effectiveness and efficacy of the proposed approach compared to SOP systems based on software-defined networking (SDN) in terms of accuracy and efficiency. The study also looks at FL enabled by SDN as a paradigm that contributes to enhancing security and minimizing training time while not compromising the learning performance. However, SDN introduces new vulnerabilities in FL, mainly DDoS attacks in the SDN control plane and attacks in the exchange of model parameters. To tackle this issue, a contrastive learning-based detector based on a two-layer deep neural architecture is introduced as a mitigation measure. In actual deployment scenarios, the LSTM SVM obtains an accuracy of 99.75% and a recall of 99.80%, whereas the CNN BiGRU obtains an accuracy of 99.36% and a recall of 99.55%.
- 30 October 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Software Engineering and Knowledge Engineering
- p. 1-16
An Energy-Efficient Cooperative DDoS Attack Mitigation Using a Federated Deep Learning Framework in SDN
A distributed denial-of-service (DDoS) attack can be launched through the huge number of edge-connected devices on software-defined AIoT systems. The lack of modern labeled training data makes it impossible to keep defending it in a centralized way, and privacy issues prevent the sharing of sensitive traffic information. In this work, a trustworthy federated learning (FL) architecture is proposed to tackle these challenges in a decentralized defense framework for AIoT. The framework incorporates three elements. First, a private model update exchange scheme based on secure multiparty computation can be used to enable privacy-preserving exchange of local model updates, while being robust to outliers. Second, a horizontal cooperation-based mitigation mechanism minimizes packet loss and decreases the effect of the attack on regular traffic. Third, a semi-supervised FL-based anomaly detection strategy is used to correctly classify attack classes and minimize the number of false alarms, based on a limited number of labeled outliers. Experimental evaluation on real-world datasets (CICIDS2017 and InSDN) demonstrates the effectiveness and efficacy of the proposed approach compared to SOP systems based on software-defined networking (SDN) in terms of accuracy and efficiency. The study also looks at FL enabled by SDN as a paradigm that contributes to enhancing security and minimizing training time while not compromising the learning performance. However, SDN introduces new vulnerabilities in FL, mainly DDoS attacks in the SDN control plane and attacks in the exchange of model parameters. To tackle this issue, a contrastive learning-based detector based on a two-layer deep neural architecture is introduced as a mitigation measure. In actual deployment scenarios, the LSTM SVM obtains an accuracy of 99.75% and a recall of 99.80%, whereas the CNN BiGRU obtains an accuracy of 99.36% and a recall of 99.55%.
- Open Access
- 29 June 2026
- Published by International Academic Institute for Science and Technology in International Academic Journal of Innovative Research
- Vol. 13 (2) , 42-49
Advances in Biotechnology for Developing Personalized Medicine Approaches in Clinical Practice
Personalized medicine has recently emerged as a revolutionary clinical paradigm that enables individualized disease prevention, diagnosis, and treatment guided by a patient's genetic, molecular, and clinical factors. While much progress has been made in applying biotechnology to realize the goals of personalized care, issues of inter-patient heterogeneity, the time lag in biomarker discovery, and the inadequacies of existing one-size-fits-all medical paradigms still undermine clinical benefit. In this study, we have examined current advances in applying biotechnology to developing clinical approaches for personalized medicine. This process integrates genomic profiling, molecular diagnostics, biomarker identification, patient stratification, and treatment selection into a single framework. We utilized 520 patient profiles from categories spanning oncology, cardiovascular diseases, neurological disorders, metabolic syndromes, and inherited diseases, and applied a structured analysis approach. Comparing the diagnostic effectiveness of the developed framework with a standard clinical protocol for disease intervention in a clinical evaluation study showed that the classification accuracy, sensitivity, and specificity of the patient characterization protocol within the biotechnology-driven framework were 94.6%, 91.8%, and 93.4%, respectively. The treatment recommendation model developed using the biomarker discovery framework achieved a drug-matching rate of 93.2%, while 87.5% of patient populations achieved clinical efficacy post-biomarker-directed therapeutic intervention. In addition, improvements of 18.7%, 16.4%, and 24.5%, and a reduction of 21.3% were observed in diagnostic accuracy, drug treatment selection efficiency, clinical treatment efficacy, and adverse drug reactions, respectively, compared with conventional clinical protocols. Our study suggests that integrating genetics and molecular analytics into precision medicine contributes to more precise diagnoses, more precise treatment decisions, and more effective therapy selection for individual patients, and will help realize its large-scale implementation as part of the next generation of clinical systems.
- 29 June 2026
- Published by International Academic Institute for Science and Technology in International Academic Journal of Innovative Research
- Vol. 13 (2) , 42-49
Open AccessAdvances in Biotechnology for Developing Personalized Medicine Approaches in Clinical Practice
Personalized medicine has recently emerged as a revolutionary clinical paradigm that enables individualized disease prevention, diagnosis, and treatment guided by a patient's genetic, molecular, and clinical factors. While much progress has been made in applying biotechnology to realize the goals of personalized care, issues of inter-patient heterogeneity, the time lag in biomarker discovery, and the inadequacies of existing one-size-fits-all medical paradigms still undermine clinical benefit. In this study, we have examined current advances in applying biotechnology to developing clinical approaches for personalized medicine. This process integrates genomic profiling, molecular diagnostics, biomarker identification, patient stratification, and treatment selection into a single framework. We utilized 520 patient profiles from categories spanning oncology, cardiovascular diseases, neurological disorders, metabolic syndromes, and inherited diseases, and applied a structured analysis approach. Comparing the diagnostic effectiveness of the developed framework with a standard clinical protocol for disease intervention in a clinical evaluation study showed that the classification accuracy, sensitivity, and specificity of the patient characterization protocol within the biotechnology-driven framework were 94.6%, 91.8%, and 93.4%, respectively. The treatment recommendation model developed using the biomarker discovery framework achieved a drug-matching rate of 93.2%, while 87.5% of patient populations achieved clinical efficacy post-biomarker-directed therapeutic intervention. In addition, improvements of 18.7%, 16.4%, and 24.5%, and a reduction of 21.3% were observed in diagnostic accuracy, drug treatment selection efficiency, clinical treatment efficacy, and adverse drug reactions, respectively, compared with conventional clinical protocols. Our study suggests that integrating genetics and molecular analytics into precision medicine contributes to more precise diagnoses, more precise treatment decisions, and more effective therapy selection for individual patients, and will help realize its large-scale implementation as part of the next generation of clinical systems.
- 8 February 2025
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 84 (30) , 37133-37166
Optimized deep transfer learning techniques for spine fracture detection using CT scan images
- 8 February 2025
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 84 (30) , 37133-37166
Optimized deep transfer learning techniques for spine fracture detection using CT scan images
- 20 September 2023
- Published by Springer Nature
- p. 513-532
Secure Group Communication Among IoT Components in Smart Cities
- 20 September 2023
- Published by Springer Nature
- p. 513-532
Secure Group Communication Among IoT Components in Smart Cities
- 23 February 2023
- Published by Springer Nature in Archives of Computational Methods in Engineering
- Vol. 30 (5) , 3113-3132
A Review on Prediction and Prognosis of the Prostate Cancer and Gleason Grading of Prostatic Carcinoma Using Deep Transfer Learning Based Approaches
- 23 February 2023
- Published by Springer Nature in Archives of Computational Methods in Engineering
- Vol. 30 (5) , 3113-3132
A Review on Prediction and Prognosis of the Prostate Cancer and Gleason Grading of Prostatic Carcinoma Using Deep Transfer Learning Based Approaches
- 10 January 2020
- Published by Springer Nature in Wireless Personal Communications
- Vol. 117 (4) , 2809-2829
Weighted Ternary Tree Approach for Secure Group Communication Among Mobile Applications
- 10 January 2020
- Published by Springer Nature in Wireless Personal Communications
- Vol. 117 (4) , 2809-2829
Weighted Ternary Tree Approach for Secure Group Communication Among Mobile Applications