Unmasking Intruders: An In-Depth Analysis of Anomaly Detection Using the KDD Cup 1999 Dataset

With the rapid development of network technologies and increasing number of complex network attacks, traditional cybersecurity approaches do not provide sufficient protection against emerging threats. In the hybrid model proposed for network anomaly detection of this paper, deep learning models like Multilayer Perceptron and Deep Neural Network are combined with classical machine learning classifiers including Decision Trees, Random Forests and Support Vector Machines. The performance of the proposed model is tested with KDD Cup 1999 dataset, which is one pf the most popular benchmark for network anomaly detection. The experiment results show that the hybrid model achieves a higher detection performance than either machine learning or deep learning approach alone. The results demonstrate that the hybrid model fuses advantageous points of diverse models, offering better performance than any other method in anomaly detectionaccuracyand robustness. This study contributes to the development of useful cybersecurity.