Networks are an essential component of modern digital infrastructure and are thus frequent targets of cyber attacks. Anomaly behavior detection has become one of the significant intrusion detection techniques, looking for abnormalities in network activity. Advanced ML algorithms, such as recurrent and convolutional neural networks, have been widely applied to optimize detection rates. Graph neural networks (GNNs) have given promising results to model complex relationships in network traffic. In this paper, authors have presented a new graph-based intrusion detection system that improves anomaly detection with GNNs. The approach improves network security by applying convolutional layers for feature extraction and edge attributes to gain more contextual information. Early experiments with the UNSW-NB15 dataset shows the credibility of the method in network anomaly detection, and therefore, it is a very viable direction for modern-day cybersecurity purposes.