This paper proposes a semi-analytical method for the time-dependent reliability assessment of steel pipelines exposed to localized corrosion. The failure modes considered in this paper include corrosion perforation and local burst due to the growth of defect depth and length. For a short pipeline, a spatial-temporal Poisson model is used to describe the occurrence of corrosion defects. The method is further generalized for the reliability assessment of a long pipeline, which is treated as a series system consisting of different segments. The spatial correlation of the behaviour of adjacent segments is taken into account, and is modeled by the Nataf transformation method. An example is presented to demonstrate the applicability of the proposed method. Analytical results show that the reliability of a short pipeline can be significantly underestimated if the assessment method for a single corrosion defect is used instead. For a long pipeline, the failure probability is smaller in the presence of a stronger spatial correlation between different segments, suggesting that the manufacturing procedure and assembling practice of different segments should be synchronized with each other in order to improve pipeline reliability.