Multi-party private set intersection enables participants to jointly compute the intersection of their datasets without disclosing any private information. However, existing approaches often suffer from high computational and communication overhead. To address this, we propose the Zero Masking Method, which enables each participant to generate a Zero Function with a special cancellation property. By integrating this method with garbled bloom filter, we design the ZMPSI protocol, which requires only a small number of public-key operations and communication rounds, thereby improving overall efficiency. We further formally prove security against up to t–2 colluding participants in the semi-honest model for both the Zero Masking Method and the ZMPSI protocol. Experiments show that, with 10 participants and a dataset size of 2 16 , the total execution time of ZMPSI is only 141 seconds, and the communication overhead is 119.57 KB for each client and 18.53 KB for server.