Water supply networks are marred by serious risks of imperceptible pipeline leakage, posing ‎sustainability ‎and ‎performance ‎threats.  This article highlights the use of vibratory signal features to get around the drawbacks of traditional methods in a highly detailed framework for leak detection based on CatBoost. demonstrated excellent diagnostic performance and carried out a thorough test performance evaluation on five leakage configurations  . The expected system achieved an accuracy of 98.1% (variance (well within x/3% of expected):, beating traditional competitors such as Random Forest (97.3%) and Support Vector Machine (93.8%). ‎For example, the area under the receiver-operating characteristic curve was 0.995, indicating perfect or near perfect discrimination. Root mean square energy (32%) and spectral entropy (25) Indeed, their diagnostic characteristic characteristics were all in line with classic fluid dynamic laws Computational ‎‎efficacy allows real-time ‎system ‎deployment, with 0 .8 ‎milliseconds ‎per ‎every ‎classification ‎mandate ‎and ‎18-‎‎megabyte ‎memory occupancy. The specifications are actionable to create compatible configurations to enable follow-up and sustainable employment of infrastructure systems. By linking recent trends in machine learning to the practice of infrastructure monitoring, this study helps bring the world a step closer to achieving the SDGs.