The advancement of automation systems increases the use of lightweight devices. An Internet of Things (IoT) network comprises lightweight devices that generate vast amounts of data through the Internet. Therefore, it is highly vulnerable to cyber-attacks and brings many challenges. It also poses risks by potentially impacting mission-critical systems like intranet and database servers. We have proposed a deep learning technique-based intrusion detection system for IoT networks. It uses deep learning techniques to classify attacks and their variants. It learns through data, including SYN, ACK, Plain UDP, UDP flood, and Scan, for analyzing network traffic patterns. This work is executed on NB-IoT and Kitsune datasets. It starts with preprocessing steps such as handling missing values, feature standardization, and dimensionality reduction, which are applied to enhance model robustness. A deep learning-based sequential architecture with dense layers, regularization, and dropout is employed to improve generalization and prevent overfitting. The proposed method accurately identifies Mirai variants, demonstrating significant improvements in precision, recall, and F1 scores. These findings underscore the potential of ML-driven real-time intrusion detection systems in proactively mitigating evolving cybersecurity threats.