Subway systems serve as vital components of urban transportation infrastructure. Due to the critical concern for transportation safety, research on the resilience and recovery capabilities of subway systems is therefore of great significance. This paper proposes a recovery model for cascading failures in subway systems and the recovery measures and cascading failures are addressed simultaneously. The recovery measures in the model include both station repairs with resource allocation and passenger evacuation. The recovery model incorporates five key factors: the maximum number of stations in repair, the sequence of station repairs, the total amount of repair resources, the allocation of repair resources, and the speed of passenger evacuation. The model is applied to different subway networks, simulating the deliberate attack on the station with the highest load. Simulation results demonstrate the superiority of the proposed capacity-based resource allocation method. Moreover, the effectiveness and importance of passenger evacuation in halting the propagation of cascading failures and improving network recovery are examined. Finally, the impact of station capacity on the recovery outcome is explored, We reveal an optimal station capacity setting that could minimize network resilience loss during attacks. This study holds significant implications for the development of resilient and sustainable subway systems.