The increasing market share of electric vehicles (EVs) makes charging facilities indispensable infrastructure for integrating EVs into the future intelligent transportation systems and smart grid. One promising facility called fast charging reservation(FCR) system was recently proposed. It allows people to reserve fast chargers ahead of time. In this system, fast chargers are the most scarce resource instead of electricity. Thus how to allocate these charging points requires careful designing. A good allocation policy should 1) ensure charging points to be allocated to EV users who really value them, and 2) prevent users’ private information, e.g., identity, personal agenda, residing area and etc., from being inferred. A simple combination of classic multi-item auction and user identity anonymization cannot satisfy both criteria simultaneously. To find such an allocation, in this paper we investigate the design of privacy-preserving auctions in FCR systems. Traditional privacy-preserving strategies such as cryptography could incur high computation and communication overhead and hence jeopardize the efficiency of allocation. To this end, we propose Auc2Reserve, a differentially private randomized auction. Auc2Reserve applies an improved approximate sampler and the belief propagation (BP) technique to accelerate the resource allocation and pricing process. As a result, it is much more computationally efficient than generic exponential differentially private mechanisms and other theoretical approximate implementations. Through theoretical analysis, we show that Auc2Reserve is ?-incentive compatible, individual rational and ?-differentially private. And it provides a close-form approximation ratio in social welfare of FCR systems. In addition, we also demonstrate the efficacy of Auc2Reserve in terms of social welfare and privacy leakage via numerical simulation.