Transport diffusion takes a major role in many processes within the chemical industry. It is the lack of experimental data for multicomponent mixtures that enforces the use of (over-) simplified theoretical models which are often, due to the complexity of the involved physics, unsuccessful. The main challenge lies in the development of accurate predictive methods for transport diffusion in fluidal multicomponent mixtures. In the recent past, molecular simulation on the basis of force fields has emerged as a viable tool for the prediction of transport properties. Molecular simulations provide a direct insight into the relationship between molecular structure and macroscopic behavior. One viable approach is to compute diffusion coefficients out of equilibrium molecular dynamics (EMD) simulations via the Green-Kubo formalism, which provides a relationship between macroscopic coefficients and the time integral of an autocorrelation function of the respective microscopic flux. Within the scope of this project Maxwell-Stefan diffusion coefficients are predicted for binary and ternary fluid mixtures. Furthermore, a methodology to compute the thermodynamic factor, in order to predict the respective Fick diffusion coefficients, will be developed. This simulation methodology will be evaluated for binary, ternary and even quaternary systems. Both Fick and Maxwell-Stefan diffusion coefficients will be compared to the available experimental data for a variety of thermodynamic states. The existing data base will be extended which allows a verification of existing classical prediction methods. Characteristics in the vicinity of the liquid binodal in the context of liquid-liquid decomposition will be researched. Moreover, a comprehensive systematic study of Lennard-Jones type model systems will be performed to gain a more fundamental insight into the dependence of the various diffusion coefficients on the thermodynamic state and the respective molecular interactions of multicomponent fluid mixtures.