This book presents some recent developments of probabilistic assessment of systems dependability based on stochastic models issued of graph theory, finite state automaton and language theory, applying as well for static as for dynamic and hybrid context. The first part of the book recalls the foundation of the coherence property represented by a state graph model and shows how the reliability of the system may be extracted from this model. An algorithm with lower complexity than classical approach (BDD decomposition for example) is proposed and extended to apply as well to non coherent systems by introducing the concept of terminal tie-set. In the second part, we introduce the model of finite state automaton to more generally represent systems and replace the concepts of cut-sets and tie-sets by the concept of event sequences. The model is enriched progressively to define hybrid stochastic automaton that allows considering all of the problems usually assembled around the concept of dynamic reliability. Examples of Monte Carlo simulations are presented.