A multi-attribute group decision making(MAGDM) method considering trust relationships in social networks was proposed to address the trust risk caused by the trust relationships among decision makers in social networks and the problem of correlation between solution attributes. Firstly, several sub-groups were obtained by clustering according to the social network relationships of decision makers, and the weights of decision makers and the weights of sub-groups were calculated by degree centrality and eigenvector centrality. Then, the trust degrees of individual decision makers and subcommunities were determined separately according to the trust relationship. The trust risk coefficient and evaluation risk coefficient were measured for the sub-group with a low trust degree. The evaluation values of the sub-group were adjusted accordingly. Secondly, the Choquet integral operator was used to obtain the collective evaluation value of the alternatives for a problem where multiple attributes of the alternatives were correlated, and the alternatives were ranked accordingly by the score function to obtain the optimal decision. Finally, the proposed decision making method was applied to an arithmetic example to illustrate the feasibility of the method.