The issue of water leakage represents a pertinent and contentious topic within the field of water supply systems (WSS). The detection and localization of such leaks is significant and pervasive challenge, primarily due to the potential to remain undetected for extended periods of time. The magnitude of water loss varies considerably, from 3% to over 50%, depending on the level of maintenance undertaken [1]. It is estimated that the global volume of water loss per year is 126 billion cubic metres, which equates to a financial loss of approximately 30 billion USD [2], and according to the Portuguese regulator ERSAR [3], the actual water leakage in Portugal's supply system in 2023 was 5.6 m3/(km day) for the bulk side and 2.4 m3/(km day) for the distribution side. To address this problem and reduce water losses, some leakage management strategies have been developed. These include preventive measures, systematic assessment protocols, regulatory frameworks, stringent control measures, detection and localization techniques, and prompt repair initiatives [4]. The detection and localization phase is of particular importance due to its inherent complexity and critical role, where a variety of techniques are employed, including hardware- and software-based methods. These techniques are also used by numerous commercial companies and have been patented in a multitude of different countries. Among the software-based techniques, machine learning models and digital twins represent valuable techniques for detecting and localizing water leaks in WSS, and they are particularly useful when considering physical laws and governing equations. Therefore, this work proposes a novel sub-framework based on machine learning techniques to detect time-series anomalies, where an anomaly represents a potential water leakage situation in the system. In this regard, different ML-based methods, including deep learning, are applied to a benchmark dataset, and compared with a baseline statistics-based methodology to detect the water leakage anomalies.