High Pressure Die Casting (HPDC), commonly used with aluminum alloys in automotive industry, is a process that produces complex components in a single block, in large quantities and in less time than other casting processes (gravity casting or lost foam casting). However, this process leads to the presence of casting imperfections resulting in poor material health. These include variations in the eutectic phase, intermetallics and the heterogeneous distribution of pores within the parts, which are very often the cause of failure. To ensure structural integrity without the need for additional weight, it is essential to consider the characteristics of defects in the fatigue design of parts. The aim of our study is to develop a fatigue design approach appropriate to the specific characteristics of HPDC materials and processes. To achieve this objective, a probabilistic Monte-Carlo type approach is developed. The synthetic materials generated are representative in terms of size distribution and spatial distribution of defects. The approach is applied to cylindrical AlSi9Cu3(Fe) specimens obtained following several manufacturing and post-processing parameters such as two cooling rate and heat treatment (as cast or partial treatment T5). X-ray tomography analyses have shown that the cooling rate is the factor that most influences the heterogeneous characteristics of the defect population in terms of size and spatial position. The internal and surface behavior models developed were applied, depending on the position of the defects. The probabilistic model also considers the competition between internal and surface defects observed with the HPDC process. It has been shown that large defects, generally of the spongy pore network type, are difficult to consider. A gaussian blurring method was used to be representative of the critical defects observed experimentally. The design of the approach enables it to be applied to real components by considering the zones of interest and applying the corresponding size distribution and spatial distribution of defects.