The proliferation of small unmanned underwater vehicles (UUVs) represents a growing security challenge in coastal waters worldwide. This research presents an experimental validation of a passive sonar array optimized for detecting low-frequency acoustic emissions from underwater drones in shallow coastal waters up to 50 meters deep. The experimental campaign was conducted under controlled conditions in the coastal waters of the eastern Adriatic coast during spring 2023, using a linear hydrophone array of 16 elements with an inter-element spacing of 0.75 meters. Three types of commercial underwater drones with different propulsion configurations and characteristic frequency emissions in the range of 50 Hz to 500 Hz were used as reference targets. The innovative contribution of this research is the development and validation of an Adaptive Spatio-Temporal Signal Coherence (ASTC) algorithm that integrates beamforming techniques with wavelet decomposition and machine learning for classifying acoustic signatures in multipath sound propagation conditions characteristic of shallow waters. Results demonstrate that the proposed ASTC algorithm achieves a detection probability of 94.2% at a false alarm rate of 2.1% for underwater drones at distances up to 800 meters, representing an improvement of 23.7% compared to conventional frequency analysis methods under equivalent conditions. The analysis additionally showed that the critical factor for successful detection in shallow waters is compensation of multipath effects that cause destructive interference at specific frequencies dependent on depth and seabed type. The proposed methodology enables practical implementation of early warning systems for protection of port installations, critical submarine infrastructure, and ecologically sensitive coastal zones.