Protecting networks against harmful URLs is becoming increasingly critical as attackers have devised new methods to spread phishing attacks, viruses, and malware through links embedded in text streams. Current works use Graph Convolutional Networks (GCNs) to capture and interpret dangerous URLs. They focus solely on the text and structure of the URL. The problem is that they employ only one type of information at a given moment and thus could miss out on crucial parts. To improve upon existing GCN models, our project incorporates various kinds of information, including content within domain names, elements of WHOIS registries, and website traffic patterns. A new technique was employed for integrating these data elements. With the use of datasets incorporating these integrated multi-modal features, we demonstrated that our method outperforms well in identifying unwanted URLs using standard metrics, such as accuracy, recall, precision, and F1 score, among others. This enhanced approach increases its reliability and robustness of malicious URL detection and also paves the way for future advancements.