The invention discloses a method for separating unknown multi-protocol mixed data frames into single protocol data frames. The method comprises the following steps that firstly, binary data are converted into a hexadecimal format, and n rows and m columns of mixed unknown protocol data frames are input; secondly, an approximate value K of type numbers of the input unknown protocol data frames is calculated; thirdly, a K value is designated by using a K-means algorithm, clustering is carried out, and n class clusters are obtained; fourthly, the good or bad of each class cluster is assessed by using a class cluster assessing algorithm based on entropy; fifthly, the class clusters with good clustering effects are placed into a result set, fingerprint information of the type is extracted, and the fingerprint information is stored in a fingerprint database. According to the method for separating the unknown multi-protocol mixed data frames into the single protocol data frames, the problems that the calculation of the approximate value of the mixed protocol type number K is difficult, the separation of different protocol areas is difficult, the assessing accuracy on protocol frame class clusters is low, and the visual and effective effects are difficult to achieve are solved.