Current Volume 9
Zero-day attacks pose significant challenges to modern cybersecurity systems due to their unpredictable nature and ability to bypass traditional detection mechanisms. Therefore, this study presents an improved machine learning-based model for the identification and management of zero-day attacks in a 5G cloud-based network. The proposed system integrated an Artificial Neural Network (ANN) model with a real-time threat isolation framework based on a novel data model made up of anomaly-based, behavioural and signature-based features. Data was collected from the Litcoder Cloud in Enugu, Nigeria and then the data was pre-processed through visualization, imputation and normalization techniques. An online adaptive feature selection mechanism using Online Recursive Feature Elimination (ORFE) was further employed to dynamically prioritize relevant threat features. The ANN model presented in this study was trained using a backpropagation algorithm with dropout regularization and evaluated using performance metrics including accuracy, precision, recall, F1-score and confusion matrix. Simulation on a 5G cloud-based environment demonstrated an average threat detection time of 0.8 seconds, high throughput of 85.10%, low CPU utilization of 0.403 and minimal latency of 27.07ms. The model achieved high classification accuracy of 98.5% and demonstrated resilience in real-time conditions, validating its capability for proactive threat detection and mitigation.
Zero-Day Attacks; Machine Learning; ANN; Threat Isolation; 5G Cloud Network
IRE Journals:
Nnadi Linda C. , Asogwa T. C.
"An Improved Machine Learning-Based Model for The Identification and Management of Zero-Day Attacks in A Cloud-Based Network" Iconic Research And Engineering Journals Volume 9 Issue 1 2025 Page 28-36
IEEE:
Nnadi Linda C. , Asogwa T. C.
"An Improved Machine Learning-Based Model for The Identification and Management of Zero-Day Attacks in A Cloud-Based Network" Iconic Research And Engineering Journals, 9(1)