The present disclosure provides an anti-malicious method, device and medium for secure three-party computation, and relates to the field of data security. The method includes the following: Respective private data matrices of three participants are determined; each participant receives a corresponding random matrix pair generated by a commodity server node; and based on the random matrix pair, a corresponding internal matrix is generated in a computational process of the three participants, and corresponding security constraints are separately added to a computational process in which a collusion behavior exists and no collusion behavior exists. The security constraints implement a constraint on a rank of an internal matrix, so that any participant in the computational process cannot predict private data matrices of another two participants. The present disclosure can improve data security of the secure three-party computation.