IoT devices are widely employed in business and industrial settings as well as daily life these days. Moreover, it will broaden the DOS attack surface. It is not possible to apply additional safety restrictions to an Internet-connected gadget because of its limited capacity. DOS attacks by malevolent actors with the ability to take over IoT devices continue to be exceedingly dangerous. We use the computer traffic network dataset in this article, which contains data with at least 50% compromised. We suggest lightweight layered LSTM-GRU for DOS attack detection. The significance of combining LSTM and GRU in DL for IoT lies in the model's ability to improve performance, efficiently handle IoT data, reduce overfitting, generate interpretable features, enhance anomaly detection, adapt to dynamic environments, and generalize across different IoT domains. The deep learning-based models that have been suggested have shown outstanding performance, with high F1 score, recall, accuracy, and precision.