A novel ensemble of hybrid starling murmuration optimized stacked gated Ghostnets for robust classification of DDoS attacks
Abstract An Internet enabled smart devices has intruded in each and every inch of individual’s habit, making them to lead more comfortable and sophisticated life. The explosive growth of these gadgets has given the bright light of many attacks that even causes the fatal end to the users. Distributed...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-08-01
|
| Series: | Discover Computing |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s10791-025-09688-x |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract An Internet enabled smart devices has intruded in each and every inch of individual’s habit, making them to lead more comfortable and sophisticated life. The explosive growth of these gadgets has given the bright light of many attacks that even causes the fatal end to the users. Distributed Denial of Service (DDoS) attacks are one such network catastrophes which causes many challenging cyber risks that arises the terrifying concerns regarding safety, security and privacy. Due to hostile user techniques and the complexity of DDoS, advanced defensive measures are mandatorily required. As the algorithms of Artificial Intelligence (AI) serves as a key factor in developing the the intelligent intrusion detection systems which accurately differentiates the known and unknown attacks. As the possible number of DDoS attacks are increasing day-by-day, an enhanced cognitive learning framework needs to be deployed to assault the DDoS attacks to the greater extent with high detection performance and less time complexity. In this regard, this research article introduces an improved framework for identifying the DDoS threats utilising the novel Hybrid Gated Attention Evoked Ghostnets with its hyper-parameters tuned by the Starling Murmuration Optimization (SMO) technique to increase the detection accuracy with the high robustness and reduced overfitting problems. The proposed ensemble learning approaches with the SMO tuned hyper-parameters yield superior detection performance. Multiple heterogeneous datasets related to the DOS attacks are utilised to analyse the proposed model utilising the metrics including precision, accuracy, specificity, F1-score and recall. Moreover, the proposed framework is statistically validated and evaluated against the other existing state-of-the-art technique. Experimental findings highlight that the proposed approach has attained nearly 100% accuracy, 100% precision, 100% recall, and a 99% F1-score, with a Wilcoxon F-statistic of 7.893 (p = 0.02), establishing it as an effective solution for identifying DDoS attacks. This study demonstrates the proposed hybrid model is reliable robust solution for dynamic DOS recognition for the multiple versions of datasets, enhancing the domain of data driven cyber hunting. |
|---|---|
| ISSN: | 2948-2992 |