Abstract: Over the past two decades, the study of model-based dependability analysis has gathered significant research interest. Different approaches have been developed to automate and address various limitations of classical dependability techniques to contend with the increasing complexity and challenges of modern safety-critical system. Two leading paradigms have emerged, one which constructs predictive system failure models from component failure models compositionally using the topology of the system. The other utilizes design models - typically state automata - to explore system behaviour through fault injection. This paper reviews a number of prominent techniques under these two paradigms, and provides an insight into their working mechanism, applicability, strengths and challenges, as well as recent developments within these fields. We also discuss the emerging trends on integrated approaches and advanced analysis capabilities. Lastly, we outline the future outlook for model-based dependability analysis.