METHOD OF EXTRACTING WEB ATTACK PATTERN BASED ON INSTRUCTION OF MACHINE LEARNING TARGET SYSTEM AND GENERATING RECONSTRUCTED PAYLOAD DATA, AND PREPROCESSOR USING SAME

摘要

Disclosed is a method of extracting a web attack pattern based on an instruction of a machine learning target system and generating reconstructed payload data comprising the steps of: (a) when payload data is obtained, performing, by a preprocessor, a process of talking to at least a part of characters, special characters, and numbers included in the payload data based on the special characters; (b) performing, by the preprocessor, a process of selecting a plurality of specific character groups each constituting a plurality of commands corresponding to a specific web attack type stored in a database; (c) performing, by the preprocessor, processes of (i) determining whether the plurality of specific character groups corresponds to each component constituting a specific command combination of the specific web attack type, and substituting each of the plurality of specific character groups with each specific first character representing each class corresponding thereto if the plurality of specific character groups correspond to each component constituting the specific command combination of the specific web attack type, and (ii) removing the plurality of specific character groups and the rest of the special characters from the payload data; and (d) performing, by the preprocessor, a process of substituting each specific first character and the special character with a real number or a second character corresponding to the real number to generate reconstructed payload data.;COPYRIGHT KIPO 2020