A novel integration of multi-stocked gated variant recurrent units and Kolmogorov-Arnold tuned deep training networks for anchoring the intrusion detection against computer attacks
Abstract With the explosive expansion of smart computers, there is a rapid intrusion of network attacks in the user’s personal life that intensifies the privacy and security breaches. An adaptive increase in the traffic communication between the smart gadgets poses the serious barriers in safeguardi...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-07-01
|
| Series: | Discover Computing |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s10791-025-09678-z |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract With the explosive expansion of smart computers, there is a rapid intrusion of network attacks in the user’s personal life that intensifies the privacy and security breaches. An adaptive increase in the traffic communication between the smart gadgets poses the serious barriers in safeguarding the gadgets against potential security breaches. Hence predicting the varied types of threats is essential for constructing the intelligent defence system against the cyber-attacks. Artificial Intelligence (AI) mechanisms notably Machine (ML) and Deep Learning (DL) models portrayed the promising performance in iidentifying and mitigating security risks aimed at these smart devices. However, these systems are facing a major bottleneck in detecting the real time datasets with the increased performance and less computational overhead. The present research work proposes the novel integration of Multi-Stocked Variant Gated Recurrent Units (MSV-GRU) and Fast Kolmogorov-Arnold based Deep Learning Networks (FKA-DLN) for the precise extraction of attributes that enhance classification of the cyber-attacks from the realistic network-based traffic data sets. The proposed framework introduces the FKA-DLN based classification network to learn the micro-level features that fuels detection performance. The major components of the proposed model: (1) Data-Collection and Pre-processing (2) Feature Extraction using MSV-GRU (3) FKA-DLN model for the better classification. The comprehensive experimentation is carried out using UNSW-2019 datasets on the proposed model and performance metrics such as precision, accuracy, specificity, F1-score and recall are examined and assessed with varied deep learning frameworks. Evaluation findings highlighted that the recommended model has attained an accuracy of 99.4%, precision of 98.8%, recall of 98.6% and F1-score of 99.2% respectively. Compared with the conventional learning frameworks, the proposed novel system has produced the better classification. |
|---|---|
| ISSN: | 2948-2992 |