1. IC Counterfeit Detection Using Physical Inspection Methods
Group 173
Project Members:
Kevin King (Electrical Engineering)
David M. Orosz (Electrical Engineering)
Christopher Smedberg (Computer Engineering)
Project Advisors:
Mohammad Tehranipoor (Dept. of Electrical and Computer Engineering)
Omer Khan (Dept. of Electrical and Computer Engineering)

2. Project Statement:
The counterfeit integrated circuit market is a major problem that can ruin company's reputations and lead to faulty products. The most common type of counterfeits are recycled IC's or IC's that have already been used but are sold as new. These used IC's may no longer function within specifications, or may not function properly.
We will take advantage of the fact that an IC that is not genuine will almost always show some physical differences in its packaging.

3. Key Concepts: IC Counterfeiting Electron Microscopy Image Processing
Blacktopping
Remarking
Re-Tinning
Electron Microscopy
Scanning Electron Microscope
Material Analysis
Image Processing
MatLab Image Libraries
Fingerprinting

4. IC Counterfeiting
BlackTopping - Resurfacing technique used to remove signs of aging
Remarking - Include blacktopping, but will be used to change type or quality of chip. Changes markings on chip
Re-Tinning - During IC recycling, damage to the pins will most likely occurr. Re-tinning attempts to restore the pins to new quality

5. Electron Microscopy
Scanning Electron Microscopes use a beam of electrons to scan the surface of something.
The interaction of the electrons with the electrons in the object being imaged, give data about topography and composition
We will use the SEM to gather highly detailed images of the IC's surface, and material analysis information of the chips capping material and pin material

6. Image Processing
When examined at a higher magnification, the difference between genuine and counterfeit chips is very evident.
This is due to sandblasting within the blacktopping process.
GENUINE
SUSPECT

7. Image Processing
These images can be analyzed using a "black space" detection process.
Genuine chip textures have a more well defined shape, and therefore less "black space"
Counterfeit chip textures have more black space
MATLAB can be used to enhance the images (so they can be compared), highlight the black spaces, and even count the amount of black pixels.
This process is demonstrated on the genuine image from the previous slide.

8. Counterfeit Detection
Test Lists and Descriptions

9. Basic Visual Test:
This test will not be automated and will use a human's best judgement
Using no magnification or very low-power magnification, the chip should be inspected for blatant defects, such as scratches or bent pins
The Lot Codes and Number of pins should be checked to make sure they match as they should

10. Marking Permanency Test
Test 1 - Using a mixture of 1 part alcohol, and 3 parts mineral spirits, use a cotton swab to rub this mixture over the surface of the chip.
A color change, removal of markings or color may indicate a counterfeit IC
Test 2 - Using Acetone, repeat the test as above. Acetone may also make a noticeable difference to the texture of the surface of a counterfeit chip

11. High Magnification Texture Test
Using a highly magnified image of the suspect chips texture, this image will be compared to a similar image of a known good chip.
Using image processing techniques, these images will be compared and it will be determined whether the image raises any questions
This test takes advantage of the fact that most blacktopping techniques will show a different texture than genuine chips

12. Some Preliminary Results
Number of highlighted spots ==>

13. Some Issues This method is not flawless
In certain areas of the chip, a large black spot may show up (even if the chip is genuine- this would skew our results.
Also, as shown in the previous slide, the amount of spots differs greatly between genuine chips (although it is MUCH higher in counterfeit)
'GENUINE1' has ~37,000 spots
'GENUINE2' has ~47,000 spots
'GENUINE3' has ~60,000 spots
With more MATLAB trials, or with the incorporation of other enhancement commands, this value could be normalized
We could focus on just the larger/smaller clumps of spots
We could only look at the outline of the shapes
etc...

14. Top and Bottom Texture Comparison
This test will be very similar to the previous texture comparison test. however, the two images that will be compared will be of the Top and Bottom of the suspect chip
Blacktopping may not be detected when comparing a suspect chip's texture to a known good chips. But comparing the top and bottom textures of the same chip may show more subtle differences.
The black spot detection process in MATLAB can also be used to show differences between the top and bottom textures. If the "number of highlighted spots" is drastically different between the two sides, the chip is suspect.

15. Pin Re-Tinning Test
Using the material analysis ability of the SEM, the material of the pin will be tested
During counterfeiting, pins will be restored using a re-tinning process.
Evidence of this process may be found when comparing the material analysis of a suspect chip to the material analysis of a known good chip

16. Surface Material Analysis
The material analysis function of the SEM will also be used to find the chemical makeup of the chips surface.
Blacktopping also uses a material different from what is used in the factories.
The differences may be subtle, depending on the quality of the counterfeit, but can be observed when comparing the material analysis of the suspect chip to a known good chip

17. Counterfeit Classification
Determining Counterfeit Type From Test Results

18. Counterfeit Taxonomy
Using data collected from the tests we will run, we hope to be able to classify any found counterfeit chips appropriately

19. Possible Defects

20. Process Automation

21. Automated Tests
Texture and Material Analysis test will run automatically once user has submitted required images
Visual Inspection and Marking Permanency tests will need to be conducted by hand

22. Program Automation
The testing process should run completely automatically, once the user has loaded the appropriate images.
The user will be responsible for an image of the top texture, an image of the bottom texture, a material analysis of the pin, and a material analysis of the top of the suspect chip.
User will also be required to submit an image of the top texture, a material analysis of a pin, and a material analysis of the top of a known good chip
Program should automatically run tests and display any tests that raised concerns of the genuinity of the chip

23. MATLAB Server Can integrate with many applications and web service's
allows matlab computations from custom applications
Can use all existing MATLAB toolboxes for image processing
Can be run remotely
Cost: $500

24. Pixel Comparing Matches each pixel in an image looking for differences
The corr2() function in MATLAB would be used to do the comparison.
Would give the most accurate results
Is highly dependent on alignment
Alignment can be done with software

25. Project Timeline

26. Budget Available : $5,000 Estimate: $1,350 Purchased: Dremel Tool $150
To-be Purchased:
De-capping service for 8051(X2) ~$150
FPGA(and decapping) ~$
Different Chip(and decapping) ~$