Crack propagation tests were conducted with CCT specimens made by 2024-T3 Aluminum alloy in benign and corrosive environments to investigate the stochastic characteristics of crack growth. A random variable probability crack growth model, base on the modified form of Paris law, was used to characterize the stochastic performance. Distribution significance test has been done, which shows that the random variable X can be considered reasonably to follow log-normal distribution in all the 4 kind of environments. Variance analysis indicates that the scatter in wet air is greater than that in laboratory air significantly. Expressions for the probability distribution of crack size at any given loading cycle and the probability distribution of the random fatigue life at which a given crack size is reached were discussed. The corresponding calculations for the crack length distribution and P-a-N curves with certain reliability were carried out. The comparisons with experimental data indicate the validity of the proposed method.