Improving Network Security with Gradient Boosting from KDD Cup Dataset

Improving Network Security with Gradient Boosting from KDD Cup Dataset

연구 분야: Analysis

학회: SN Computer Science

Network traffic analysis plays a crucial role in detecting and mitigating security threats in modern computer networks. In this study, we employ machine learning techniques, specifically Gradient Boosting, Linear Discriminant Analysis (LDA), and Support Vector Machines (SVMs), to analyze network traffic data from the KDD Cup dataset. The KDD Cup dataset contains a large volume of network traffic data, including various features such as duration, protocol type, service, and flag, among others. We first preprocess the dataset by removing redundant rows and converting categorical variables to factors. We then split the dataset into training and testing sets and perform feature scaling to ensure uniformity across features. For our analysis, we utilize Gradient Boosting, a powerful ensemble learning technique, to iteratively train weak learners to correct the errors of the previous models. Additionally, we employ Linear Discriminant Analysis to find the linear combination of features that best separates the classes, and SVMs with both linear and Radial Basis Function (RBF) kernels for classification. We evaluate the performance of each model using standard metrics such as accuracy, precision, recall, and F1-score. Furthermore, we visualize the loss percentage over iterations for Gradient Boosting to gain insights into the optimization process. Our results demonstrate the effectiveness of Gradient Boosting, LDA, and SVMs in accurately classifying network traffic instances. The models show promising performance in distinguishing between normal and malicious network activity, highlighting their potential for enhancing network security in real-world applications.