ELECTRONIC COMPONENT CLASSIFICATION

• US 20180024178A1
• Filed: 09/06/2017
• Published: 01/25/2018
• Est. Priority Date: 12/13/2013
• Status: Active Grant

• Current Assignee: Battelle Memorial Institute
• Original Assignee: Battelle Memorial Institute
• Inventors: House, Larry J., Engelhart, Dale C.
• Granted Patent: US 10,054,624 B2
• Time in Patent Office: Days
• CPC Class Codes: G01R 29/26 Measuring noise figure; Measuring signal-to-noise ratio
• CPC Class Codes: G06F 18/24147 Distances to closest patterns, e.g. nearest neighbour classification
• CPC Class Codes: G06F 21/44 Program or device authentication
• CPC Class Codes: G06F 2218/12 Classification; Matching
• CPC Class Codes: G06V 20/80 Recognising image objects characterised by unique random patterns

Abstract

A system and method of electronic component authentication or component classification can reduce the vulnerability of systems (e.g., satellites, weapons, critical infrastructure, aerospace, automotive, medical systems) to counterfeits. Intrinsic deterministically random property data can be obtained from a set of authentic electronic components, processed, and clustered to create a classifier that can distinguish whether an unknown electronic component is authentic or counterfeit.

Claims

1. A method of classifying an unknown electronic component comprising:

temporarily installing a plurality of known electronic components into a component interface of an electronic component authentication system for signal measurement;

measuring, during an idle state, a noise signal for each of the plurality of known electronic components via the component interface;

building a classifier based on the noise signals of the plurality of known electronic components; and

classifying the unknown electronic component using the classifier.

2. The method of claim 1 wherein building a classifier includes:

segmenting each noise signal into a noise vector;

transforming each noise vector into a feature vector;

conducting a statistical analysis on each feature vector; and

clustering the feature vectors to create the classifier.

3. The method of claim 2 wherein the feature vector is time independent and spatially independent.

4. The method of claim 2 wherein the feature vector is in the frequency domain.

5. The method of claim 2 wherein said conducting the statistical analysis on the feature vector includes organizing the feature vector by variance and discarding dimensions where the variance is below a threshold.

6. The method of claim 2 wherein said conducting the statistical analysis on the feature vector includes organizing the feature vector by variance and discarding all but a predefined number of dimensions that have the highest variance.

7. The method of claim 2 wherein said clustering includes selecting a clustering algorithm based on properties of said unknown electronic component.

8. The method of claim 2 wherein said clustering includes comparing the feature vectors to each other using a distance metric.

9. The method of claim 2 wherein transforming each noise vector into a feature vector using a hybrid transformation, wherein said hybrid transformation includes transforming each noise vector using a plurality of different transformations that are combined to create each feature vector.

10. The method of claim 1 wherein each of the plurality of known electronic components belongs to one of a plurality of chip classes.

11. The method of claim 1 wherein measuring a noise signal for each of the plurality of known electronic components includes measuring a power consumption signal for each of the plurality of known electronic components.

12. The method of claim 1 wherein each noise signal is a deterministic intrinsic noise signal unique to that electronic component.

13. The method of claim 1 including measuring, during an idle state, a noise signal of the unknown electronic component via the component interface and wherein classifying the unknown electronic component includes classifying the unknown electronic component using the classifier and the noise signal of the unknown electronic component.

14. The method of claim 1 wherein the plurality of known of electronic components are temporarily installed one at a time into the component interface of the electronic component authentication system.

15. The method of claim 1 including processing the noise signal of each known electronic component to generate a noise signature of each of the plurality of known electronic components;

and wherein said building a classifier based on the noise signals of the plurality of known electronic components includes building the classifier based on the noise signatures of the plurality of known electronic components.

16. A method of classifying a plurality of electronic components comprising:

temporarily installing a plurality of electronic components into a component interface of an electronic component authentication system for signal measurement;

measuring, during an idle state, a noise signal for each electronic component via the component interface; and

classifying the plurality of electronic components into classes based on similarity of the noise signals.

17. The method of claim 16 wherein measuring the noise signal for each electronic component includes measuring a power consumption signal.

18. The method of claim 16 including building a classifier based on the noise signals of the electronic components.

19. The method of claim 16 wherein at least two of the classes represent different generations of electronic components.

20. The method of claim 16 including processing the noise signal of each electronic component to generate a noise signature of each of the plurality of electronic components;

and wherein said classifying the plurality of electronic components into classes based on similarity of the noise signals includes classifying the plurality of electronic components into classes based on similarity of the noise signatures.

21. The method of claim 20 wherein each noise signature is a deterministic intrinsic noise signature unique to that electronic component.

22. The method of claim 20 including:

segmenting each noise signature into an intrinsic deterministically random noise vector;

transforming each intrinsic deterministically random noise vector into a feature vector;

conducting a statistical analysis on each feature vector; and

clustering the feature vectors to create the different classes.

23. The method of claim 22 wherein the feature vector is time independent and spatially independent.

24. The method of claim 22 wherein the feature vector is in the frequency domain.

25. The method of claim 22 wherein said conducting the statistical analysis on the feature vector includes organizing the feature vector by variance and discarding dimensions where the variance is below a threshold.

26. The method of claim 25 wherein said conducting the statistical analysis on the feature vector includes organizing the feature vector by variance and discarding all but a predefined number of dimensions that have the highest variance.

27. The method of claim 22 wherein said clustering includes comparing the feature vectors to each other using a distance metric.

28. The method of claim 22 including transforming each noise vector into a feature vector using a hybrid transformation, wherein said hybrid transformation includes transforming each noise vector using a plurality of different transformations that are combined to create each feature vector.

29. The method of claim 16 wherein the plurality of electronic components are temporarily installed one at a time into the component interface of the electronic component authentication system.