Abstract: Due to the low memory resources and hidden dangers of information security in IoT devices, this paper proposes a lightweight ResLSTM intrusion detection system for detecting intrusion attacks in IoT devices and protecting the information security of IoT devices. This intrusion detection system incorporated the depthwise separable convolutional structure into ResNet and then combined it with LSTM, which utilized the residuals and depth- separable structure to improve the network performance and computational capability, saved computational resources, and extracted the spatial and temporal features of the anomalous traffic for data identification and classification. The final test results on the datasets UNSW- NB15, NSL- KDD, and CIC- IDS2017 show that the ResLSTM model is able to effectively detect the attack data, and the model has good generalization and robustness, with superior detection results.