Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{184046,
        author = {Balakrishnan S and Giriprashaath M},
        title = {Deep ensemble learning with pruning for ddos attack},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {4},
        pages = {12-18},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=184046},
        abstract = {With growing size and complexity of Internet of Things (IoT) networks, Distributed Denial of Service (DDoS) attacks can be induced to them, which can lead to substantial network performance degradation and disruption of services. The traditional DDoS detection systems, which are founded on static mechanisms such as threshold-based or signature-based detection, are not able to cope with new patterns of attacks. These systems are plagued with high false positives and response latency because they are designed on the basis of past data and pre-configured rules that are not updated in real-time. Therefore, such systems are not able to effectively detect new and advanced attack methods, leading to response delays and disrupted legitimate traffic. In order to overcome these limitations, the suggested solution applies an ensemble-based deep learning architecture that learns in real time from real-time network traffic data. The ensemble architecture uses VGG19 and RESNET50, both convolutional neural networks (CNNs), which have proven to be highly successful in several image recognition and pattern detection problems. Using these algorithms, the model can detect patterns of malicious traffic in real time and update its detection methods as new forms of attacks emerge. In contrast to static conventional models, the solution improves the adaptive nature of the system, which is well equipped to detect new and emerging DDoS attack methods.
Further, the system's scalability allows it to accommodate bigger and more complicated network environments without any loss in performance. With increasing IoT networks and data generated, the system can adapt to the higher growth while retaining its high accuracy of detection. Both of the ensemble models (VGG19 and RESNET50) have their own strengths, including VGG19's detection of fine image-like features and RESNET50's depth of detection of complex patterns, creating a stronger detection mechanism that pools the complementary strengths of the two models.
A key benefit of the system in its proposed state is its capability for integration with IoT network programmability. Integration into programmable ability of the IoT networks provides a mechanism where real-time re-direction of valid traffic is undertaken amidst an existing DDoS attack, leaving the services functional and untouched and inhibiting the malicious traffic. The feature will come in handy especially for critical mission IoT use cases, like the healthcare industry, industrial control systems, and smart cities, whose downtime or loss of services has critical effects.
Generally, the system as proposed provides a strong counter to DDoS attacks in current dynamic IoT scenarios. It promises quicker decision-making, enhanced accuracy in detection, minimized false positives, and streamlined rerouting of traffic, effectively enhancing network robustness and the security of IoT networks. The solution is ideally fit for the requirements of today's IoT systems, where adaptability and scalability in real time are critical.},
        keywords = {DDoS detection, Internet Of Things (IOT), VGG19, RESNET50 (CNN), ensemble learning, real-time mitigation, network security, traffic analysis, adaptive DEEP Learning.},
        month = {August},
        }