The fast growth of smart homes, which is made possible by adding IoT devices, has made life in a house a lot easier and more automated.  Although this makes it easier to connect to the internet, it also opens up a lot of security holes that let hackers and other bad people into homes.  In this study, we suggest a new way to make smart homes safer by using Graph Neural Networks (GNNs) to find intrusions.  The main goal of this method is to use the complicated connections and relationships between the different smart devices in the home network to find strange behaviour that could be a sign of a security threat. By representing the smart home network as a graph with devices as nodes and their interactions as edges, GNNs can detect local as well as global patterns of device activity. Intrusiveness detection systems therefore become more accurate and efficient. In our sense, we construct a live graph-based model of the smart home environment that illustrates the gadget communication and information sharing. GNNs examine these graphs and learn to identify deviations from usual patterns of interaction. For instance, indicators of an intruder's presence may include illegal access or malfunctioning devices. We investigate the proposed approach using a real-world smart home dataset and demonstrate that GNNs can effectively identify unusual activity across a broad spectrum of devices, including thermostats, security cameras, and door sensors. According to the results, GNNs outperform popular machine learning techniques such decision trees and support vector machines in terms of object identification and false positive generation. This work demonstrates how robust, versatile, and real-time systems for smart homes able to detect intruders might be produced using GNNs. It also makes it possible to look into how graph-based models can be used to improve security in other IoT-based settings. This work promotes the creation of better and more reliable smart living areas by making it easier for smart home systems to spot intrusions on their own.