In this paper, a new meta-heuristic algorithm inspired by human behaviors and the population-based optimizer, called special forces algorithm (SFA) is proposed based on the situation of special forces performing tasks in the real environment. In order to effectively simulate the characteristics of special forces, the SFA introduces different tactical behaviors and group strategies in reality into the optimization ideas, and designs a unique search mode. Special forces perform three tasks according to specific scenarios and mission requirements: large-scale search and assault search, capture and rescue. By combining the different strategies, and adding some uniquely mechanisms to the algorithm, the SFA simulates real dynamic behaviors to meet optimization requirements. The proposed SFA is compared with other types of mature algorithms, the performance of the SFA is verified in 15 sets of benchmark function tests including the unimodal function, multimodal function and fixed-dimensional function. The results demonstrate that the SFA has shown great potential and competitive results. The SFA can obtain good search performance and optimization accuracy on the basis of a better balance of exploration and exploitation capabilities.