Cyberattacks on critical Internet of Things (IoT) systems must be detected using Intrusion Detection Systems (IDS). It is necessary to enhance the protection of these devices from cyberattacks in order to safeguard their users. A common feature of IDSs' anomaly detection components is deep learning (DL) techniques.ML approaches have been used for many years to increase the reliability and resilience of Network Intrusion Detection Systems (NIDS). In the last few years, to identify intrusions in the Internet of Things (IOT), deep learning algorithms are employed. Intrusion is one of the major issues in IOT due to lack of security, attacks on network, accuracy etc., in the network. In order to overcome this problem, a novel Hybrid Convolutional neural network and Graph convolutional network (CNN-GCN) technique has been proposed in this work for IOT intrusion detection. The proposed CNN-GCN deals with attacks issue on network, accuracy rate and lack of security in IOT. The proposed work improves feature extraction, accuracy and attention-based classification. Performance evaluation of the suggested method will be done using the MATLAB simulator. The proposed technique performance has been assessed using specific metrics such as accuracy, classification and error detection. In comparison to the existing methods, the findings demonstrate that the proposed work exhibits 99.86% error detection in IOT.