Authors
Faculty of Engineering, University of Science and Technology, Sudan
[email protected]
Abstract
Cyber Security is a smart defence strategy that uses advanced artificial intelligence as a shield. Artificial Intelligence (AI) and Machine Learning (ML) enhance cyber security by enabling faster and more accurate threat detection, automated responses and adaptive defence system that learn from data. Network security on the other hand, involves protecting, monitoring and controlling the network infrastructure and the data transmitted across it.
In this study predictive analysis was performed to classify network threats based on key performance metrics such as band width, latency response time and jitter. The network data set consisted of 1000 instances and 5 attributes to evaluate the classification performance; several data mining algorithms namely Naive Bayes, IBK, SMO, Logistic, and Random Tree were implemented using the WEKA data mining tool.
The experimental results indicate that the Random Tree classified achieved the best performance compared to the other algorithms. The accuracy obtained by the five algorithms was as follows (Random Tree 99.8%) IBK (95.9%) Naive Bayes (94.8%) Logistic Regression (87.9%) and SMO (87.7%). These findings demonstrate that the Random Tree model provides superior predictive capability for network threat detection. The confusion matrix analysis indicates that the proposed model has a strong capability to correctly identify attack traffic. With minimal misclassification. The model also demonstrates reliable performance in distinguishing between warning and normal network states. These results confirm it effectiveness in detecting real time performance degradation and identifying potential network threats. The evaluation process includes measuring model accuracy to determine the effectiveness of the classification results. The findings indicate that machine learning based approaches significantly improve network monitoring performance and enhance cyber security decision making processes. These results demonstrate the importance of predictive models in identifying network anomalies and supporting proactive threat mitigation strategies.