Focusing on the issues that information acquired by different participants are not equal in the Private Set Intersection (PSI) protocol based on Garbled Bloom Filter (GBF), which can not be effectively applied to the cloud environment, a cloud outsourcing PSI protocol combined the garbled Bloom filter algorithm with the proxy oblivious transfer protocol was proposed. Firstly, by introducing the garbled Bloom filter, the problem of false positive in the traditional standard Bloom filter was solved to achieve efficient storage and large data transmission. Secondly, the complex time-consuming computation could be outsourced to the cloud proxy server by using proxy oblivious transfer protocol, so that the cloud tenants did not need to be online in real-time and only needed a small amount of computation. Finally, in the processing of the cloud outsourcing privacy set intersection, the comparison results could be fairly obtained without the interaction among the cloud tenants. Theoretical analysis and performance comparison show that the communication and computation complexities of the proposed protocol are linear, and the proposed protocol is safe and effective.