Protein-protein interfaces, which are regions of interaction between two protein molecules, contain information about patterns of interacting functional groups. Recognition of such patterns is useful both for prediction of binding partners and for the development of drugs that can interfere with the formation of the protein-protein complex. We present a novel method, Interface-to- interface (I2I)-SiteEngine, for structural alignment between two protein-protein interfaces. The method simultaneously aligns two pairs of binding sites that constitute an interface. The method is based on recognition of similarity of physico-chemical properties and shapes. It assumes no similarity of sequences or folds of the proteins that comprise the interfaces. Similarities between interfaces recognized by I2I-SiteEngine provide an insight into the interactions that are essential for the formation of the complex and can be related to its function. Its high efficiency makes it suitable for large scale database searches and classifications.