Initiation and growth of pitting corrosion results from electro-chemical reactions closely related to environmental factors, causing great uncertainties in the pit depth, pit size, pit shape and pitting distribution. A cellular automaton (CA) approach was adopted to address the electro-chemical process of pitting corrosion in marine environments, yielding a geometric morphology of corrosion damage. The simulated corrosion pits in the CA based model, treated as the cylinders, semi-ellipsoids, cones or cavities of hybrid shapes, were introduced into structural analysis models. The effect of pit shape on compressive behavior and ultimate strength of pitted plates was studied, which was linked with the stress interaction between pits. A modelling method based on stochastic simulation was proposed to generate random pitting damage based on limited statistical corrosion data, which could reflect all aspects of the random natures of actual pitting corrosion. Numerical results showed high consistencies with available tests of ultimate strength of pitted plates. The pit shape significantly affects the ultimate strength of pitted structures, likely causing a strength variation more than 20% in a naturally pitted hull plate. This is essentially due to the stress interaction between pits of diverse shapes affecting the onset and development of plasticity in pitted areas.