Vision Transformers for Counterfeit IC Detection

The global electronic semiconductor supply chain is a complex network that is essential for modern technology. However, this system faces serious challenges, particularly the risk of counterfeit components entering the market, which can undermine product quality and reliability. The widespread distribution of manufacturing and distribution processes makes it difficult to maintain consistent quality and increases the chances of fraud. With the advancements in deep neural networks, modeling solutions can be developed to combat the infiltration of counterfeit ICs in the market. In this paper we examine how the global nature of the semiconductor supply chain contributes to the problem of counterfeit parts and propose a vision transformer model to accurately classify between authentic samples and 11 different types of counterfeit defects. Our model achieves a remarkable $88 \%$ accuracy on test set. The attention maps from the vision transformer are also visualized to get insights into the predictions from the model.