A numerical technique for identifying the elastic properties of an adhesive layer of composite materials from experimental data is presented. The problem is posed as the “inverse homogenization problem”. Homogenization problem was solved by the asymptotic averaging method. A wide literature review of papers devoted to solving and investigating the problem of inverse homogenization is presented. Finite element method was used for solving “local problem” on periodic cell of composite. The inverse problem was reduced to the optimization problem, for which the method of sequential quadratic programming was applied. Tikhonov’s regularization with a classical apriori approach to the choice of the regularization coefficient was used for regularize the solution. The method of homogenization taking into account the adhesive layer of the composite material is considered. The adhesive layer is introduced as an additional isotropic phase of the composite. The Monte Carlo method was used to analyze the stability of the identification problem solution to the errors in the experimental data. Dispersed-reinforced composite filled with glass microspheres is considered in computational experiments. A comparison with the experimental data is given. The necessity of including of an adhesive layer in model for calculating the effective elastic characteristics of dispersed-reinforced composite materials is substantiated. Approach of the elastic properties of the adhesive layer identification from experimental data taking into account the noise of mathematical models and experimental errors is approved.