Optical fiber sensor networks (OFSNs) are facing the problem of a lack of systematic evaluation criteria to assess network performance. In this paper, a universal quantitative robustness evaluation model for OFSNs is proposed. The model defines robustness as the mathematical expectation of the monitoring coverage ratio, which has taken into account the performance under all possible network states and the probability of each state. This model is applied to four basic network topologies including line, ring, star and bus topologies, and their mathematical expressions of robustness are derived by analyzing all possible states in detail. Further simulation gives a quantitative comparison among these topologies, proving that the ring and star topologies are optimal for the monitoring of strip-shaped and square regions, respectively. Finally, two influencing factors, the attenuation coefficient and the threshold, are investigated for their impact on the robustness of the network.