Nowadays, deep-submicron CMOS technologies are basic for the development of modern computer-based systems, whose use simplify our everyday life in a variety of domains such as e-government, electronic commerce and banking and terrestrial and aerospace transportation. The steady reduction of transistors size allows for less power consumption and higher frequency rates, leading altogether to greater performance. These same features that improve VLSI systems performance, negatively affect their dependability. Low-power, high-speed, reduced-size transistors are highly increasing the likelihood of occurrence and the diversity of faults affecting these systems. Therefore, there exists a great interest in developing new and efficient techniques to assess the dependability of deep-submicron manufactured systems in the presence of faults. Fault injection, which consists in the deliberate introduction of faults into a system, is a well-known approach for coping with this problem. In this context, modelbased fault injection has the interest of enabling the dependability assessment of the system in the early stages of its development cycle, thus reducing the costs of fixing any error. However, the long time required to simulate large and complex models make it impractical in many cases. This thesis focuses on the use of Field-Programmable Gate Arrays (FPGAs) to accelerate model-based fault injection experiments by implementing the model of the system under study on reconfigurable hardware. It improves existing research on FPGA-based fault injection in two different directions: i) it studies existing semiconductor technologies to determine a representative set of transient and permanent fault models to cope with, and analyses to what extent such models, can be emulated by means of FPGAs, and ii) it determines, for each considered fault model, alternative procedures to carry out its emulation, evaluating the resulting speed-up. FADES (FPGA-based Framework for the Assessment of the Dependability of Embedded Systems) is the prototype that has been developed in order to illustrate the feasibility of the approach. The experimentation deployed using FADES is mainly focused on validating the correctness of the results provided by the proposed prototype and showing the attainable speed-up of the solution when compared with a state of the art simulation-based fault injection tool. Finally, a discussion reflecting the main advantages and drawbacks of the proposed approach and its usefulness for the dependability assessment of embedded systems is provided.