Cybercriminals are interested in the Internet of Things (IoT) more than ever due to its remarkable growth pace. This assertion is supported by the growing number of cyberattacks on IoT devices and intermediary communication mediums. IoT attacks that go unnoticed for a long time can result in serious service disruptions and monetary losses. Real-time intrusion detection on IoT devices is necessary to ensure the dependability, security, and profitability of IoT-enabled services. This study proposes a lightweight deep-learning method for detecting intrusions in IoT devices. The proposed system uses a hybrid Convolution Neural Network (CNN) with Long Short Term Memory (LSTM). Three distinct models, CNN, LSTM, and the proposed hybrid CNN+LSTM were used to identify intrusions in the UNSW-NB15 dataset. The proposed hybrid model was used to identify event characteristics on a Raspberry Pi3 device. To minimize computation costs, CNN and LSTM were stacked without the final layer to maximize convergence speed. CNN and LSTM layers are mapped to the sample marker space using fully linked layers and a softmax layer. The average accuracy, recall, precision, and F1-score of the proposed hybrid intrusion detection system were 98.78%, 98.09%, 97.88%, and 97.99%, respectively.