A method for identifying a network application. The method includes analyzing metadata and source code of a network application to extract a set of application tokens, generating an index document of the network application based on the set of application code tokens, wherein the index document is included in a library of index documents corresponding to a number of network applications, extracting a set of packet header tokens from a packet header of a packet in a flow, comparing the set of packet header tokens to the set of index documents to generate a number of match scores, wherein each match score represents a similarity measure between the set of packet header tokens and one index document, and determining, based on a highest match score corresponding to a particular network application, that the flow is generated by the particular network application.