Optimization of Hyperparameter Tuning for Attack Detection in IoT Networks Using a Deep Belief Network

Eko Arip Winanto (Universitas Dinamika Bangsa)
M. Riza Pahlevi B (Universitas Dinamika Bangsa)
Dodi Sandra (Universitas Dinamika Bangsa)
Kurniabudi (Universitas Dinamika Bangsa)

PROCESSOR Jurnal Ilmiah Sistem Informasi, Teknologi Informasi dan Sistem Komputer
Vol 21 No 2 (2026): Jurnal Ilmiah Processor, Edisi Oktober 2026

An Intrusion Detection System (IDS) is a crucial tool in computer network security for detecting adversarial attacks. Deep Learning is a trending method in both research and industry, and it can recognise subtle differences in data. However, the effectiveness of an IDS depends on the accuracy and efficiency of the detection model employed. Additionally, deep learning models require careful tuning of hyperparameters to achieve optimal detection performance. This study explores the impact of hyperparameter tuning on the performance of a Deep Belief Network (DBN) in detecting network attacks. Two experimental scenarios were conducted: DBN without hyperparameter tuning and DBN with hyperparameter tuning. The results indicate that the DBN with hyperparameter tuning achieved an accuracy of 0.999, outperforming the DBN without hyperparameter tuning (0.998 accuracy). These findings suggest that hyperparameter optimisation significantly enhances the detection capability of the model, improving both the accuracy and the resilience of the IDS against attacks.