The Internet of Things (IoT) has transformed modern technology by connecting smart gadgets. While these advancements provide unparalleled potential, they also pose complicated security issues. Deep Learning has shown potential for identifying and detecting cyberattacks on IoT devices. Intrusion Detection System (IDS) is critical for protecting sensitive data by detecting and mitigating suspicious activity in the IoT setting. To overcome these issues a novel Deep Learning-Based Cyber-attack detection in IOT device (CDDI-IDS) has been proposed in this paper, for effectively detecting and categorizing the types of attacks in IoT. The process starts with collecting data from IoT devices, which is then transferred through the pre-processing stage from the IOT device, which records network data and log files. The processed data is subsequently cleaned and normalized to reduce noise and improve model performance. Renyi Entropy is used for the feature extraction process to select the most important feature while minimizing noise. Dingo Optimization is used to tune hyperparameters, improving accuracy and efficiency. Finally, the trained model was classified using ConvBiGRU, which detects sequential patterns in time-series data as Normal, Probe, R2L, DoS, and U2R. This deep learning-based technique improves cyber threat detection accuracy, making it perfect for protecting IoT networks from hostile activity. Experiments on the KDDCup99 datasets demonstrate that intrusion detection system outperforms existing approaches and models. The CDDI-ID method achieves 99.23% accuracy, while the DNN, LSTM, CNN methods achieve 96.56%, 92.2%, and 94.5%, respectively