Crowdsourcing has given rise to spatial crowdsourcing, which has been widely used in many fields such as air quality data collection, smart city construction, natural disaster or traffic status reporting. However, spatial crowdsourcing faces two major challenges. Since workers need to disclose real-time location information to complete tasks at a specific location, the risk of location privacy leakage is high, which will further expose sensitive information such as medical records and home addresses. Secondly, it is still extremely difficult to strike a balance between the accuracy of multi-dimensional feature matching and privacy protection considering factors such as spatial location, skills, and reward range. To address these issues, we propose a multi-dimensional feature privacy-preserving matching scheme for the first time in this work. Specifically, an innovative homomorphic comparison protocol for Jaccard similarity calculation is designed, and a dynamically searchable encryption structure and multiple privacy-related protocols are proposed. Extensive experiments on real datasets show that the scheme can effectively balance privacy protection and task matching accuracy, and also performs well in terms of privacy level, matching time, and structure construction time.