Human cortical folding is highly convoluted and characterizes the inter-subject variability. Recent studies found that the adult brain cortex is unique for individual identification. However, little is known about whether the infant brain cortex, which develops dynamically in the first postnatal years, is reliable for individual identification. To this end, we proposed a novel morphological folding descriptor, called FoldingPrint, to perform the infant identification in a large longitudinal dataset with 472 infants. Successful identification results indicate the effectiveness of the proposed FoldingPrint. In addition, we found that the regions with high identification accuracy are mainly distributed in high-order association cortices.