[Objective]In view of the significant uncertainty in geotechnical parameters in deep excavation engineering and the difficulty of traditional site investigation data in accurately reflecting the actual soil conditions after excavation, this study proposes a Bayesian inversion and dynamic reliability updating method for geotechnical parameters by integrating monitoring data. [Methods]By introducing a Bayesian updating framework combined with Markov Chain Monte Carlo (MCMC) sampling, displacement monitoring data of the excavation are utilized to dynamically invert and update the probability distributions of soil cohesion and internal friction angle, effectively reducing parameter variability. On this basis, a quadratic response surface surrogate model is employed to replace time-consuming numerical simulations, and Monte Carlo simulation is conducted to efficiently evaluate excavation reliability, forming an integrated analysis framework of “monitoring-parameter inversion-reliability assessment.” A deep excavation project in Huaihua City, Hunan Province, is taken as a case study to validate the proposed method. [Results]The results indicate that, after incorporating monitoring data, the standard deviations of the posterior distributions of soil parameters are significantly reduced, leading to a notable decrease in geotechnical parameter uncertainty. The failure probability of the excavation calculated based on the updated parameters is less than 10−6, which is much lower than the prior value of 1.9×10−5, and the reliability index is significantly improved, demonstrating that the excavation is in a favorable safety condition. [Conclusion]The proposed method can more realistically reflect the actual safety state of deep excavations and provides an effective tool for risk management and safety assessment during excavation construction.