Backup paths are widely used in IP networks to protect IP links from failures. However, existing solutions such as the commonly used independent model and Shared Risk Link Group (SRLG) model do not accurately reflect the correlation between IP link failures, and thus may not choose reliable backup paths. A cross-layer approach is proposed for minimizing routing disruption caused by IP link failures. A Probabilistically Correlated Failure (PCF) model is developed to quantify the impact of IP link failure on the reliability of backup paths. With the PCF model, an algorithm is proposed to choose multiple reliable backup paths to protect each IP link. When an IP link fails, its traffic is split onto multiple backup paths to ensure that the rerouted traffic load on each IP link does not exceed the usable bandwidth.