Addressing supply chain risks of microelectronic devices through computer vision

Abstract

Microelectronics are at the heart of nearly all modern devices, ranging from small embedded integrated circuits (ICs) inside household products to complex microprocessors that power critical infrastructure systems. Devices often consist of numerous ICs from a variety of different manufacturers and procured through different vendors, all of whom may be trusted to varying degrees. Ensuring the quality, safety, and security of these components is a critical challenge. One possible solution is to use automated imaging techniques to check devices' physical appearance against known reference models in order to detect counterfeit or malicious components. This analysis can be performed at both a macro level (i.e., ensuring that the packaging of the IC appears legitimate and undamaged) and a micro level (i.e., comparing microscopic, transistor-level imagery of the circuit itself to detect suspicious deviations from a reference model). This latter analysis in particular is very challenging, considering that modern devices can contain billions of transistors. In this paper, we review the problem of microelectronics counterfeiting, discuss the potential application of computer vision to microelectronics inspection, present initial results, and recommend directions for future work.