Abstract. For a network system survivability refers to the ability to provide essential services to end users in the presence of failures and/or attacks. Survivability evaluation focuses on the measurement of this a-bility. As a dominating way of survivability evaluation, the model-based analysis technology for wireless sensor networks(WSNs) faces three prob-lems. First, the technology assumes that node distribution accords with some statistical regularity, and network has a fixed topology. However, in WSNs the node distribution and topology may change at any time which makes the evaluation result be untrustworthy. Second, survivabil-ity indicators are calculated by hand. But the personnel calculation is error prone. Third, there is no a systematic way to compute survivabil-ity indicators, i.e., we are not able to perform calculations of indicators by calling fundamental calculation processes. To solve the first problem, we propose a Continuous Time Markov Chain(CTMC) to characterize evolution of behaviors of single nodes under failures and attacks, and a parallel composition of CTMCs to characterize evolution of behaviors of WSN. The parallel composition can characterize links between nodes such that the calculation of survivability indicators does not depend on the node distribution regularity and topology. To solve the second prob-lem, we develop an algorithm to map the initial deployment of WSN to the parallel composition of CTMCs described by the modeling language of the stochastic model checker PRISM. To solve the third problem, we present how to formalize survivability indicators including k-connectivity etc. with Continuous Stochastic Logic(CSL). Finally, we show the effects of node misbehaviors on survivability by numerical analysis.