Uncertainties are ubiquitous in various engineering and science fields. Examples include analysis and design of structures and infrastructure systems against natural or manmade hazards, rocket and satellite design in aerospace engineering and safety analysis in nuclear engineering. To enhance the performance of those systems, actions taken by designers and decision-makers should be toward a set of performance objectives with higher reliability or resilience. Moreover, as sensing technologies are maturing and becoming more cost efficient, allowing their implementation at large scales, information about the state of the built and natural environments is becoming more available. This information can be leveraged to reevaluate or update forecasts of the performance of these systems and enhance confidence in our forecasts of the future performance. Analysis of the new information can therefore lead to more effective risk-informed decisions. Uncertainty Quantification (UQ) techniques such as reliability analysis and updating can help with quantitative assessment and real-time updates of infrastructure performance through the estimation of probability of failing to meet one or a set of objectives. The state-of-the-art techniques based on surrogate models, such as Kriging, open new avenues for reliability analysis by adaptively learning the shape of the limit state and substituting the originally time-consuming performance function with the estimated one. However, (I) the process of unnecessary training, (II) lack of accuracy measurement for the failure probability estimate, (III) high computational demand for high-dimensional problems, and (IV) lack of capability to perform reliability analysis with real-time updating still remain as significant challenges. To address the aforementioned limitations, this study offers the following novel contributions:-A methodology called Reliability analysis through Error rate-based Adaptive Kriging (REAK) is proposed to significantly reduce the number of calls to sophisticated, computationally demanding models and precisely control the error of estimated probability of failure.-The theoretical confidence intervals for the probability of failure estimated through the adaptive Kriging-based reliability analysis is mathematically derived.-For reliability problems with high dimensionalities, a novel algorithm is proposed by deeply integrating the Kriging surrogate model and subset simulation.-A new reliability updating framework based on Kriging surrogate model is proposed to enable real-time reliability updating.