The study explores the critical issue of man-made cavities beneath structural foundations, highlighting their potential to jeopardize stability and cause extensive damage. By examining a range of factors such as soil characteristics, operational variables, and void attributes, the research aims to elucidate their collective influence on foundation-bearing capacity and settlement. Of particular concern are the implications of subsurface voids in large-scale infrastructure like pipelines and tunnels, which can undermine their operational integrity. To address the reliability of such structures, a novel approach that combines ANN-based surrogate modeling with FORM analysis was developed. For this, 272 datasets were extracted using Python-based automation in PLAXIS 2D by considering the foundation, soil, and cavity parameters. Leveraging a hybrid metaheuristic strategy that integrates dragonfly optimization (DFO) and particle swarm optimization (PSO), the study enhances the efficacy of the surrogate model. The hybrid model of ANN-DFO displays excellent performance in terms of a high R2 value of 0.987, a low root mean square error (RMSE) of 1.22, and a very low mean absolute error (MAE) of 1.65. The simulated noise varied with specific variance (p) from 0.01 to 0.5, thus making it possible to check the efficacy of the model at different white Gaussian noise levels. The settlement equation derived from this advanced model is used to conduct FORM analysis, and hence, valuable insights into the reliability of the foundation under uncertain conditions are gained.