MACHINE-LEARNING BASED DETECTION METHOD AND APPARATUS OF FIELD PROGRAMMABLE GATE ARRAY HARDWARE MALICIOUS FUNCTION TRIGGER

摘要

The present invention provides a machine-learning based detection method of field programmable gate array (FPGA) hardware malicious function trigger. The detection method of FPGA hardware malicious function trigger includes: a look-up table (LUT) learning step of classifying only the LUTs executing a trigger function in malicious hardware function samples, and extracting featured information to enable the learning of LUTs including malicious hardware functions; a specific information extracting and learning model input step of extracting programmable interconnect point (PIP) and LUT information from target XDLs, and extracting specific information to input the same into a learned model; and a malicious hardware function detecting step of classifying the new data output from the learned model in accordance with the similarity with a learning data set, to detect the malicious hardware function. Accordingly, the machine-learning based detection method of FPGA hardware malicious function trigger can be used to detect malicious functions and verify FPGA chips.;COPYRIGHT KIPO 2019