EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Hybrid Network Model Using Advanced Learning for Detection of DDoS Attack in IoT Network

Vivek Kumar, Rajendra Kumar Dwivedi, Shiva Prakash

Computer and Decision Making: An International Journal · 08.11.2026 · 3 (8S) · 257-268

DOI: 10.1922/comdem.v3i8s.19161 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 08.11.2026. Verfasst von Vivek Kumar und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The heterogeneous network is subjected to Distributed Denial-of-Service (DDoS) attacks, and intelligent detection, cost-sensitive security decisions are necessary.Intelligent detection and cost-sensitive security decisions are needed for the increasing deployment of Internet of Things (IoT) devices in heterogeneous networks. The study suggests a Hybrid Network Model (HNM) to combine the feature selection based on mutual-information, Deep Neural Network (DNN) classification, Reliability-weighted learning, and Adaptive Prediction and Detection (APD). The proposed architecture consists of three hidden layers with 128, 64 and 32 neurons, respectively, whereas the APD mechanism converts the predicted attack probability into an allow, monitor, or block action in order to minimize the expected security costs. A scenario based analytical evaluation takes into account 6,120 network traffic samples, which is a combination of 6,000 benign and 120 malicious samples. The above Confusion matrix results in an accuracy of 99.24%, attack precision of 76.19%, recall of 88.89% and F1-score of 82.05%. The APD scenario results in 88.89% attack alert recall, 77.78% attack blocking proportion and 0.22% false blocking proportion. The three-action policy has an explicit cost of 39 units, while allowing traffic would require a cost of 180 units, which is 78.33% lower. A missed attack and false blocking cost sensitivity analysis also illustrates the effects of missed attack and false blocking costs on security responses. The results shown here are scenario calculations and not experimental. The proposed framework provides a structured approach to integrating the probabilistic attack identification into the operational decision making. The predictive reliability, computational efficiency and real-life effectiveness of DDoS mitigation in resource constricted IoT networks should be measured in future empirical validation on CIC-DDoS2019 and CSE-CIC-IDS2018.

Zitieren

APA

Kumar, V., Dwivedi, R. K., & Prakash, S. (2026). Hybrid Network Model Using Advanced Learning for Detection of DDoS Attack in IoT Network. Computer and Decision Making: An International Journal, 3(8S), 257-268. https://doi.org/10.1922/comdem.v3i8s.19161

Vancouver

Kumar V, Dwivedi RK, Prakash S. Hybrid Network Model Using Advanced Learning for Detection of DDoS Attack in IoT Network. Computer and Decision Making: An International Journal. 2026;3(8S):257-268. doi:10.1922/comdem.v3i8s.19161

BibTeX
@article{Kumar2026,
  author = {Kumar, Vivek and Dwivedi, Rajendra Kumar and Prakash, Shiva},
  title = {Hybrid Network Model Using Advanced Learning for Detection of DDoS Attack in IoT Network},
  journal = {Computer and Decision Making: An International Journal},
  year = {2026},
  volume = {3},
  number = {8S},
  pages = {257--268},
  doi = {10.1922/comdem.v3i8s.19161}
}
Kostenlos starten →