Improving security through efficient attack detection systems becomes increasingly vital in light of the growing IoT. This research evaluates the effectiveness of two DL models, CNN architecture and LSTM architecture, in detecting attacks within IoT environments. The CNN model shows remarkable accuracy, achieving 100% precision for detecting normal activities and 100% recall for identifying attacks, indicating almost flawless detection with minimal errors. However, there is a significant imbalance between precision and recall, especially in attack detection. On the other hand, the LSTM model, while slightly less accurate with a 94% accuracy rate, offers a more balanced performance with precision and recall around 95% in both categories. This balance suggests that LSTM may provide more reliable and consistent detection for IoT security. Thus, while CNN excels in accuracy and precision, the LSTM's balanced approach between precision and recall offers a more reliable option for effective attack identification in IoT systems.