Enhancing Network Intrusion Detection using Artificial Neural Networks: An Analysis of the UNSW-NB15 Dataset

Securing networks from illicit activities has remained the top priority in the constantly evolving world on information technology. Intrusion detection systems are struggling to keep up with the increasing complexity of network attacks which increased the demand for advanced methods such as Machine Learning and Deep Learning. This work emphasizes the usage of Artificial Neural Networks (ANN) for enhancing the skill to detect network incursion in IDS. Leveraging the data compilation UNSW-NB15, which evokes various internet traffic and threat instances, we created ANN-based IDS that takes benefit from neural network to detect specific patterns in data. To train and evaluate the model, multiple optimizers such as SGD, RMSprop, and Adam, in two configurations i.e. with and without dropout were compared. The reliability of newly built IDS is ensured by substantial data preparation, feature selection, along with the model tuning. The UNSW-NB15 dataset, well-known for its high data volume and accurate network inclusion, provides a significant standard to assess the impact of the suggested framework. The study shows that the ANN model using Adam optimizer and dropout outperforms with an accuracy of 94.81%, demonstrating the potential to enhance the safety of networks greatly. This paper highlights the productivity of ANNs in improving the practice of detection of network attacks, giving a solid method to combat the increasing cyber risk through extensive evaluation and testing.