Internet of Things (IoT) has transformed how people interact with the physical environment, generating vast amounts of data from interconnected devices and sensors. This study offers a comprehensive survey of IoT network traffic datasets. The efficient analysis of these massive datasets has become a challenging task, leading researchers and practitioners to harness the power of machine learning algorithms. This paper delves into the exciting field of IoT dataset analysis using machine learning techniques. We explore the different IoT datasets on which ML is implemented, offering insights into the future prospects of this rapidly evolving domain. Overall, we examine seventeen datasets that we found through the analysis of various scientific publications, which we have covered in this study. Additionally, we examined which ML approach is more effective at classifying attacks and how we can implement Feature Engineering or Data Pre-processing before applying ML techniques on these datasets. Many researchers get some limitations while applying the ML techniques on different datasets which we have discussed in our analysis, also they highlights emerging trends and future directions on these datasets. Atlast, we have discussed five well known datasets which has been created and given labels corresponding to their different attack categories.