Traffic　identification　using　statistic　features　of　traffic　flows　has　attracted　a　great　deal　of　interest.　One　challenging　issue　for　these　methods　is　that　they　ignore　the　short　flows　containing　just　2-3　packets,　for　statistic　features　of　short　flows　are　mainly　insignificant　or　meaningless.　A　topology-based　method　combining　heuristic　rules　and　deep-in　packet　inspection　is　proposed　to　identify　the　application　types　of　traffic　flows.　The　experiment　results　demonstrate　that　the　method　can　get　higher　precisions　and　similar　recalls　compared　to　deep-in　packet　inspection　method.©　(2012)　Trans　Tech　Publications,　Switzerland.