Multi-task assignment is a critical challenge in spatial crowdsourcing (SC). Compared to single-task allocation, Multi-task allocation facilitates a more efficient and effective distribution of resources in spatial crowdsourcing, thereby enhancing task execution efficiency and data quality. However, multi-task allocation has the problem of privacy leakage. Existing privacy-preserving multi-task allocation schemes rely on noise addition or homomorphic encryption (HE) to protect privacy, but this brings problems of inaccurate allocation or high computational overhead. To overcome these limitations, we propose EPOMTA, an efficient and privacy-preserving online multi-task allocation scheme based on secret sharing and somewhat homomorphic encryption (SHE). EPOMTA consists of three core components: a secure computation module, a utility computation module, and a multi-constraint task allocation algorithm. We prove EPOMTA’s security guarantees and validate its computational efficiency and allocation effectiveness through extensive experiments against state-of-the-art approaches.