In the ever-evolving landscape of network security, effective intrusion detection is paramount.The detection of anomalies in network systems is very vital to the safety of data as well asthe privacy of individuals. This research presents an advanced approach to network intrusiondetection by incorporating Graph Convolutional Networks (GCN). We propose two novel mod-els: a GCN-Autoencoder (GCN-AE) for unsupervised anomaly detection and a GCN-basedk-Nearest Neighbors (GCN-KNN) for supervised intrusion classification. Utilizing a balanceddataset extracted from the UNSW-NB15 dataset, which includes a diverse range of modern net-work activities and cyber-attacks, our models exploit the underlying graph structure of networkdata to enhance detection capabilities. The imbalanced dataset consists of 49 features and 700,000instances including standard network traffic metrics such as duration, protocol type, service,source bytes, and destination bytes which help to gain deeper insights into complex attacks. TheGCN-Auto-encoder performed at 52.86% whereas the GCN-KNN performed at 99.94% with thebalanced dataset.