Utilizing neural networks with CICIDS2018 dataset for detecting brute force attack anomalies in intrusion detection systems

Ahmad Heryanto (Universitas Sriwijaya, Indonesia)
Adi Hermansyah (Universitas Sriwijaya, Indonesia)
Triwanda Septian (Universitas Sriwijaya, Indonesia)
Ali Bardadi (Universitas Sriwijaya, Indonesia)

Jurnal Mantik
Vol. 7 No. 4 (2024): February: Manajemen, Teknologi Informatika dan Komunikasi (Mantik)

In this study, the effectiveness of neural networks in Intrusion Detection Systems (IDS) has been tested using the CICIDS2018 dataset to achieve accurate intrusion detection results. The research findings reveal that several neural network parameters will reach optimal results with a learning rate of 0.1, a training and testing data proportion of 80:20, and an optimal number of nodes in the hidden layer of 4. Other parameters such as a minimum error of 0.0001 and 2500 iterations also play a crucial role in improving IDS capability. Based on the research, it is shown that neural network models can provide optimal results in detecting intrusion patterns. This study can assist in the development of reliable and efficient neural network-based IDS to address the challenges of intrusion detection