As the Internet of Things (IoT) rapidly proliferate in the world, new attacks exploiting the weaknesses of the unfledged IoT technologies are emerging constantly. An Intrusion Detection System (IDS) is a powerful tool to defend IoT systems against security threats by monitoring abnormal activities on networks. As an effective approach to detecting malicious behaviors, Machine Learning (ML) has gained substantial interest from researchers. An ML-based IDS framework for IoT systems is proposed in this study and ten learning methods are applied for performance evaluation based on a recently published dataset, the TON_ IoT network dataset. Experimental results show that the stacking-ensemble model is the most optimal classifier, obtaining Matthews correlation coefficient (MCC) scores of 0.9971 and 0.9909 in the binary classification and the multiclass classification, respectively.