This work proposes the use of the advanced neural architecture of the Feature Tokenizer (FT)-Transformer for the Intrusion Detection System (IDS) in an IoT environment. By benefiting from the powerful self-attention in transformers, the FT-Transformer captures and identify complex and complicated dependencies and interactions among features in IoT data. We conducted a series of of experiments to evaluate the proposed TF-Transformer for assessing and enhancing. The RT_IOT2022 dataset used in training and evaluating the proposed model. The performance of the model is assessed based on the resulting metrics of accuracy, precision, recall, and F1-score. The experimental results showed that the FT-Transformer improved the performance of cyberattack detection in an IoT network and, in comparison to Deep Learning (DL) models such as CNN, RNN, and autoencoder, could offer high accuracy and robustness in output prediction. Results were found which indicated that FT-Transformer model could have a potential application to improve IoT security and provide robust frameworks for further research and development.