To address the network attack risks caused by security vulnerabilities in Internet of Things (IoT) devices

an efficient intrusion detection system (IDS) is designed to enhance protection.Accordingly

a deep learning model combining convolutional neural network(CNN) and long short-term memory (LSTM) is proposed. CNN extracts spatial features from IoT traffic for pattern recognition

while LSTM analyzes the temporal dependencies of traffic data to achieve binary classification of malicious and benign activities. After training and testing on the CICIoT2023 dataset and further validation on the CICIDS2017 dataset

the results show that the model achieves an accuracy of 98.42%

a loss value of 0.0275

a false positive rate (FPR) of 9.17%

and an F1 score of 98.57%.It is concluded that the proposed CNN-LSTM model effectively improves the accuracy and efficiency of intrusion detection

significantly enhancing the security protection capabilities of IoT systems against network threats.