In Wireless Sensor Networks (WSNs) instead of using the possible network connectivity to its maximum extent, a deliberate choice must be made to restrict the topology of the network. Constructing a virtual backbone network using Connected Dominating Sets (CDS) is a promising choice for topology control. Currently, almost all existing studies employ heuristic and/or meta-heuristic optimizations for formulating minimum-sized CDS under the deterministic network model. In this paper, we address the problem of constructing energy efficient CDS in WSNs while improving network reliability. The problem is modelled as a multi-objective optimization that simultaneously maximizes two contradictory parameters: reliability and energy efficiency. Unlike most of the existing studies, the reliability parameter is expressed as a probabilistic inference using probabilistic network model due to uncertainty in connections among sensor nodes. Extensive simulation results indicate that the proposed approach in this paper achieves more reliability, longer stability period and more energy efficient CDS compared to other approaches in the literature.