1. University of Connecticut Automated Counterfeit IC Physical Defect Characterization Team 176 Wesley Stevens Dan Guerrera Ryan Nesbit Advisors: Professor Mohammad Tehranipoor Professor Domenic Forte Electrical and Computer Engineering

2. All Rights Reserved 2 Summary Automated system for identifying physical defects Analyze various images (Optical, X-Ray, SEM) Return relevant data regarding counterfeit status

3. All Rights Reserved 3 Background Threat of counterfeit ICs increasing Over 1 million counterfeit ICs found in military supplies Can cause critical failure of systems Leads to loss of life in military and medical applications Current physical defect analysis done manually Need expert to spend time on tests Tests can be destructive Subject to human error

4. All Rights Reserved 4 Image Capture Several GB of optical and SEM images obtained already Needs to be consistent in terms of lighting as well as distance from lens to chip Image capture of the top, bottom, and sides of the chip Different algorithms can be used for specific parts of chip

5. All Rights Reserved 5 Project Overview

6. All Rights Reserved 6 Project Overview Three main steps Acquire images of suspect and/or “golden” ICs Run different algorithms based on image location Algorithms return altered images with highlighted defects Uses images from different locations to find defects Leads/Pins – scratches, bends, corrosion Surface – scratches, discoloration, pattern variation Markings – missing, faded, different location

7. All Rights Reserved 7 Image Groupings Golden-IC Analysis Take identically positioned images for one golden IC and one suspect IC Use comparison algorithm to determine inconsistencies Self-Reference Analysis Take images from different locations of the package of a suspect IC Use comparison algorithm to determine inconsistencies Group Comparison Analysis Compare data of individual ICs to group average Large variation suggests counterfeit

8. All Rights Reserved 8 Example Group Comparison

9. All Rights Reserved 9 Algorithms Binary Transformations Text recognition Ghost markings, extraneous markings Edge detection Feature acquisition and measurement Statistical Analysis Texture comparison Scratches, color variation, different pattern, corrosion, contamination, package damage Feature matching Image alignment

10. All Rights Reserved 10 Binary Transformations

11. All Rights Reserved 11 Feature Extraction & Edge Detection One key goal is the ability to recognize defects. This can be achieved through feature extraction. Allows us to look at individual objects. If it is a part of the device we can measure and compare it to other objects If it is not then we can try to identify what it is. Also allows us to delve into text recognition.

12. All Rights Reserved 12 Statistical Analysis

13. All Rights Reserved 13 Feature Matching

14. All Rights Reserved 14 Project Plan Finish algorithms for finding each defect in taxonomy Combine algorithms into one cohesive program Automate algorithm ordering and parameters Refine outputs to give meaningful results for user