Fatigue damage and atmospheric corrosion lead to the degradation of cable performance, which affects the lifetime safety risk of the cable-stayed bridge. In order to investigate the influence of cable resistance degradation on the system reliability of cable-stayed bridges, a series-parallel probability model of the cable resistance was established. A framework for time-variant system reliability evaluation of cable-stayed bridges was presented based on machine learning. Both a typical bridge and a long-span cable-stayed bridge were selected as prototypes to investigate the influence of the cable degradation on their structural system reliability indices. Numerical results show that the dominant failure sequence of the short-span bridge with short-spacing cables changes from the bending failure of girders and pylons to the tensile failure of cables due to the cable corrosion. As a result, the structural system reliability significantly decreases in the period that the cable reliability is inferior to the one of the critical girder. For the long-span bridge with long-spacing cables, the system reliability index decreases to the threshold value 5.2 in the 29-year service period taking into account both fatigue and corrosion.