Disclosed is a core triangle-based local community discovery method applicable to a social network. The method comprises the following steps: 1) a core triangle selection phase: a, determining a core node; b, finding all triangles formed by the core node and neighbor nodes thereof; c, selecting a core triangle having the highest degree; d, ending; 2) a community expansion phase: a, taking the core triangle as the initial community; b, calculating the node adaptability of neighbor nodes of each community; c, adding the neighbor nodes having the largest node adaptability to the communities to iteratively generate local communities; d, ending; and 3) a local community merging phase; a, calculating the similarity between two local communities; b, merging every two local communities based on a similarity threshold; c, ending. The method does not need to acquire global network structure information, but can generate local communities of the social network only by iteratively acquiring local information, thereby providing a solution for discovering a local community structure in the social network.