Aiming at the problems of traditional chip detection methods，such as low detection efficiency，high requirements and poor applicability，a chip detection framework based on the electromagnetic side channel and ma? chine learning method is proposed．Firstly，on the basis of holding the genuine chip，the feature vector is extracted by introducing the neural network and various feature extraction methods，and the instruction signal of the positive sample is used as the template library. Then，the near-field electromagnetic signals of the chip to be tested are di? vided into frames，and the features of each frame are extracted. Finally，the feature vectors are input to One-Class Support Vector Machine with improved kernel function for scanning matching，so as to achieve the purpose of chip detection. The experimental results show that the proposed method can be applied to the detection of counterfeit chips that have been re-marked such as shoddy chips and fake chips.