The invention discloses a malicious code detection method and system based on multi-feature fusion. The detection method includes the steps that firstly, N-gram and variable-length N-gram of code sequences in a training set and a test set are extracted to serve as a training set basic feature set and a test set basic feature set; secondly, a training set effective feature set and a test set effective feature set are screened out from the extracted training set basic feature set and the test set basic feature set respectively through an information gain mode; thirdly, the screened training set effective feature set is subjected to feature fusion through the Boosting algorithm to obtain a fusion feature classifier; fourthly, by means of the fusion feature classifier, the screened test set effective feature set is detected. By means of the detection method, finally obtained fusion features have higher identification performance and robustness, and malicious codes can be better classified out.