This paper discusses modern model-based design aspects for ensuring high dependability of mechatronic systems, i.e. ensuring most reliable and safe operation under presence of non-avoidable threats. An introductory assessment clarifies relevant terms of reference such as 'systems' (in particular mechatronic systems), 'models', 'design' and 'dependability' with special focus on the effect of threats (faults, errors, failures). The further considerations give answers to the questions 'What 'dependability' models (methods) have to be used?' and 'How to work with these 'dependability' models (methods)?' in the context of building dependable systems that are robust against threats. Results of current research at the TU Dresden Automation Engineering Lab demonstrate the successful applicability of model-based system threat analysis to control systems for robotic vehicles introducing new concepts such as dual graph error propagation model, error propagation for hybrid block diagram and finite state machine models, error propagation in multi-rate time discrete models, optimized software-implemented fault tolerance and model-based selective regression testing.