The present application discloses a method and an apparatus for file identification. The method for file identification comprises: determining a virus family of each malicious file sample in a plurality of the file samples resulting in a plurality of virus families; dividing the plurality of the virus families into at least one sample group based on a number of the malicious files belonging to each of the plurality of virus families; training the malicious file samples in each of the at least one sample group with a different training rule to obtain at least one file identification model; and determining, using the at least one identification model whether a file is a malicious file. The method for file identification of the present application may provide different identification models for various types of malicious files and thus improves the accuracy of the file identification.