The Internet of Things (IoT) is an ever-expanding network of interconnected devices that enables various applications, such as smart homes, smart cities, and industrial automation. However, with the proliferation of IoT devices, security risks have increased significantly, making it necessary to develop effective intrusion detection systems (IDS) for IoT networks. In this paper, we propose an efficient IDS for complex IoT environments based on convolutional neural networks (CNNs). Our approach uses IoT traffics as input to our CNN architecture to capture representational knowledge required to discriminate different forms of attacks. Our system achieves high accuracy and low false positive rates, even in the presence of complex and dynamic network traffic patterns. We evaluate the performance of our system using public datasets and compare it with other cutting-edge IDS approaches. Our results show that the proposed system outperforms the other approaches in terms of accuracy and false positive rates. The proposed IDS can enhance the security of IoT networks and protect them against various types of cyber-attacks.