In recent decades, Wireless sensor networks have made significant strides, drawing interest from the scientific and industrial communities. Such networks' scattered sensor nodes function autonomously in challenging environments, leaving them susceptible to mistakes and attacks that could reduce the reliability and accuracy of sensor readings. Sensor readings are categorized as aberrant data, outliers/anomalies when they considerably vary from the predicted healthy behaviors. Such outliers can have a significant influence on the decision-making process and subsequent results in data analytics. As a result, the academic community has recognized the use of machine learning algorithms for outlier detection in WSNs as an innovative and promising methodology. On the basis of numerous viewpoints taken from the body of current research, we present a thorough definition of outliers in this work. We offer a novel and creative method to identify sensor irregularities by utilizing machine learning techniques. By utilizing pattern recognition and anomaly detection methods, machine learning enables us to analyze sensor data and find outliers. We give a comparative assessment of several approaches using machine learning paradigms for outlier detection in WSNs in order to provide a thorough understanding. For academics and practitioners looking to choose the best strategies for their unique application settings, this overview is an invaluable resource. In the end, we explore the main issues surrounding the identification of outliers in WSNs. The dynamic nature of WSNs, the finite resources of sensor nodes, the changing climatic conditions, and the requirement for real-time detection are only a few of the problems that these difficulties cover.