This paper presents the development and evaluation of a lightweight and interpretable intrusion detection system tailored for IoT environments. Using the Random Forest algorithm, the model successfully distinguishes between normal traffic and various attack types such as brute force, DDoS, ARP spoofing, and Man-in-the-Middle (MitM), even in resource-constrained settings. The model was trained and tested using both real and synthetic traffic data collected from IoT environments. Performance analysis reveals 100% precision, recall, and F1-score for all classes, with attention paid to the risk of overfitting due to class imbalance. The feature importance analysis shows that time-based and ARP-related attributes played dominant roles in the detection process. This study demonstrates that Random Forest provides an optimal trade-off between interpretability, robustness, and low computational demand for IoT network traffic analysis.