Counterfeit Parts Defect Characterization (Internal Project 2013)
The migration from a vertical to horizontal business model has made it easier to introduce counterfeit electronic components into the supply chain. Aside from the billions of dollars lost due to counterfeiting, counterfeit parts are often below specification and/or of substandard quality and, therefore, create risks for the life-critical systems and infrastructures that incorporate them. Put simply, with hardware operating at the lowest abstraction level, an unreliable hardware component can create reliability and security issues even at higher abstraction levels (software, network, etc.). While the risks and economic impacts have prompted more research in counterfeit detection, this area still suffers from many issues (i) Large overlap between existing tests that leads to higher than necessary overheads (ii) Lack of coordination and consistency among test labs; (iii) Lack of automation and reliance on subject matter experts; (iv) Failure to adapt tests and develop new ones that evolve with counterfeiters. In this project, we shall address the above issues. First, we will develop a flow/infrastructure that captures all the data on counterfeit parts from multiple sources (in real-time), stores it in a central database, and exploits advanced metrics and real-time data to dynamically improve the accuracy and overheads of counterfeit detection. Second, we will investigate how machine learning algorithms can be used to automate the counterfeit detection process and discover new defects. This project is ambitious, novel, and very valuable to the community at large (government, industry, and academia). Ultimately, the work we perform in this project will improve the efficiency and effectiveness of counterfeit detection and avoidance as well as create new opportunities in the fight against counterfeiting.
Oct. 1, 2013 – Oct. 1, 2016
Year 1 (Due Oct. 1, 2014): We will develop a new reporting format that improves the consistency of counterfeit defect data reported to the Government Industry Data Exchange Program (GIDEP). We will acquire the tools/equipment needed to inspect defects and use them to build a database with images and electrical measurements from counterfeit parts. Our results will be used to update test confidence metrics and select the subset of tests resulting in lowest cost.
Year 2 (Due Oct. 1, 2015): We will develop machine learning algorithms for optical, TeraHz, and X-ray imaging that extract part features/defects and use them to automatically classify a suspect part as authentic or counterfeit. We will collaborate with the G-19 test committee to validate our algorithms on active and obsolete parts.
Year 3 (Due Oct. 1, 2016): We will build a database containing measurements from authentic parts and develop machine learning algorithms that mine the counterfeit and authentic component databases to discover new defect patterns. A final report on all the machine learning algorithms and their associated results will be delivered to the CHASE consortia member companies.
Prof. D. Forte. Assistant Professor, ECE Department, UCONN
Prof. M. Anwar. Professor, ECE Department, UCONN
Prof. M. Tehranipoor. Charles H. Kanpp Associate Professor, ECE Department, UCONN
Halit Dogan, PhD Student, ECE Department, UCONN
- Advanced Physical Inspection Methods for Counterfeit Detection, International Symposium for Testing and Failure Analysis (ISTFA), 2014
- A Comprehensive Framework for Counterfeit Defect Coverage Analysis and Detection Assessment, Journal of Electronic Testing: Theory and Applications (JETTA), 2014.
- Counterfeit Integrated Circuits: Detection, Avoidance, and the Challenges Ahead, Journal of Electronic Testing: Theory and Applications (JETTA), 2014