Electronics authentication using electrical measurements and machine learning

The problem of counterfeiting in electronics is not recent but still critical today. Identifying counterfeit devices can be a complex task since not all suspicious items are necessarily inauthentic. The paper deals with the non-destructive detection of counterfeiting in electronics by using only electrical measurements. This approach paves the way for machine learning classification-assisted counterfeit detection through electrical measurements. Physical de-processing provides the final confirmation. • 4 electrical features from 551 (originals and fakes) LMxxx were extracted. • ML algorithms were exploited to distinguish between original and fake devices. • KNN and linear SVM models were trained and tested on the collected dataset. • The best performing algorithms were validated on various types of LMxxx. • The destructive removal of the packages confirmed the classification results.