<![CDATA[The goal of this study is to enhance the classification accuracy of fake bandwidth using a CNN model, leveraging network logs collected in real-time. For this research, the network logs from the Cyber Security Laboratory of the University of Technology Sarawak are used as a dataset for training the CNN model. The dataset consists of 20 days of continuous network activity logging, which results in over 500,000 data entries. According to the model evaluation results, the trained CNN model demonstrated high accuracy in classifying genuine bandwidth (Precision: 0.92, Recall: 0.95). Moreover, it achieved considerable success in detecting fake bandwidth (Precision: 0.89, Recall: 0.90) and the no heavy activity category (Precision: 0.98, Recall: 0.84). Analysis of Loss Over Epochs showed a dramatic decrease in loss during the training phase, with optimal convergence reached by epoch 2000. Identifying these characteristics enables monitoring systems to classify network data with high certainty, detecting bandwidth manipulation in expansive networks. Thus, this research aids the design of dynamic network monitoring systems that require minimal response time while maintaining high accuracy.]]>
Copyrights © 2025
