Basin-scale groundwater security assessments are crucial for understanding the interconnectedness of water availability, sustainability, and broader environmental and socio-economic systems. The research was conducted in the semi-arid, hard-rock terrain of the North Koel River Basin, India, where water scarcity is critical during the dry season. It introduces a holistic approach that integrates geophysical methods, SWAT hydrological modelling, machine learning, and deep learning techniques to assess groundwater security at the basin scale. Eighteen thematic datasets were analysed to identify key influencing factors, including physiographic variables, blue and green water indicators, groundwater demand-related parameters, and water quality index. Four predictive models- Artificial Neural Networks (ANN), Support Vector Machines (SVM), Random Forest (RF), and Convolutional Neural Networks (CNN) were applied to classify groundwater security into high, moderate, low, and critical zones. The CNN model demonstrated the highest accuracy with an AUC score of 0.920, outperforming RF (0.897), SVM (0.854), and ANN (0.821). The CNN-predicted groundwater security map was validated using Electrical Resistivity Tomography (ERT) surveys at four sites, confirming shallow aquifer conditions in highsecurity zones and deeper, less accessible aquifers in critical zones. This framework offers a robust, scalable tool for basin-scale groundwater security assessment and resource management in similar hydrological regions.