Intrusion Detection Systems (IDS) utilize machine learning to improve their performance in monitoring computer net-works' potential attacks. It is important to understand that the KDD Cup 99 dataset contains simulation data of computer network attack forms. However, because of the complexity of this data, ma-chine learning is required to determine the form of attacks. Artifi-cial Neural Network (ANN) and Random Forest (RF) techniques to classify forms of attacks against machine learning-based Intrusion Detection Systems. However, the performance of ANN and RF is still not optimal with Lower Quartile value below “0.5” for ANN and RF. As a result, we suggest Boundary Seeking Generative Ad-versarial Network (BGAN) to improve performance of ANN and RF in detecting attacks. Results indicate that BGAN can improve ANN and RF performance.
Ramli Ahmad, Li-Hua Li, Alok Kumar Sharma
2022 IEEE International Conference on Consumer Electronics - Taiwan