Reassessing Fingerprint Evidence

A report in New Scientist magazine informs us of an important breakthrough made in identifying fingerprints (Wade, 2024). Fingerprint analysis is based on the fact that no two fingerprints are alike. For many years, since Dr Henry Faulds designed his classification system, the evidence from fingerprints was considered a gold standard in determining whether or not a person had been present at a crime scene. However, doubts have been raised about the quality of fingerprint analysis. With recent advances, now is a good time for reassessing fingerprint evidence.

How is fingerprint analysis performed?

The Crime Scene Investigator will locate all evidence for collection. After the marks are taken from the scene, using methods discussed in this article from 2021, an analyst will assess whether the image is clear enough for comparison. A print may be smudged or there might not be enough of the print left behind on a surface to provide proper identification. However, if the analyst decides to go ahead they will then make a note of the features of the mark. These include not only the whorls and arches but also minutiae such as deltas and islands as shown in image 1. Any scars or other features are also noted at this point. This happens before any comparison with a sample print.

Image 1: Fingerprint analysis requires identification of all the features.
Credit: https://commons.wikimedia.org/wiki/File:Lukisan_sidik_jari.jpg

Only once all the features are identified can the finger mark be compared with a sample print of a known individual. This individual can be a person of interest or an elimination subject. Otherwise, a sample print can come from a database of fingerprints kept by law enforcement agencies.

A side-by-side analysis is performed, often with a comparison microscope which shows both images together, see image 2. The analyst carefully examines both marks, again noting all the features. Using a point-counting method of similar characteristics, the conclusion can be either a match, not a match, or inconclusive.

Image 2: A Comparison microscope is used to compare fingerprints.
Credit: https://commons.wikimedia.org/wiki/File:ComparisonMicroscope.png

After the first analyst gives their opinion, a second examiner is required to verify the conclusion (GFJC, 2013).

Doubts about the quality of Fingerprint analysis

From the above, you can realize that a match is mostly formed as an opinion of the analyst. In the UK it requires at least 16 points of similarity to declare a match, but different countries apply different standards. No one has provided a mathematical basis for deciding how many people in the world might share a fingerprint characteristic. Nor has there been any study on the error rate of potential matches (Mnookin, 2003). Since the conviction of Shirley McKie based on fingerprint was overturned, doubts about the validity of fingerprint analysis have crept in.

Shirley McKie, a Scottish police officer, was accused of leaving her fingerprints at the scene of a crime. As her assigned role required that she remain outside the scene, then this was a serious offense. She denied the charge repeatedly, stating that she had not entered the scene.  On this basis, she was arrested and tried for perjury. It took additional experts, provided by the defense, to prove that the print did not belong to her hand and clear her of the charge. By then her police career was ended (O’Neill, 2006).

So, for fingerprints to retain their validity, new methods of analysis must be researched.

Enter AI fingerprint analysis

Last year’s controversy lay in the field of ‘Artificial Intelligence’ (AI). In their rush to outdo artists and other creators, the programmers forgot that AI’s true utility is as a tool to help people work more efficiently. Art, by its very nature, is not meant to be efficient. It’s supposed to be an outpouring, not a calculation. However, hidden underneath all this shouting, some interesting advances have occurred in forensic science.

AI excels at pattern recognition. As described above, fingerprint analysis is all about pattern recognition. New Scientist reports on a new technique that may just return confidence in this forensic technique that has provided invaluable evidence for police. Scientists at Columbia University in New York trained an AI, using over 5000 fingerprints from 150 people.

Image 3: The Device of Columbia University.
Credit: Public Domain found at https://commons.wikimedia.org/wiki/File:Columbia_University_Shield_Light_Blue.svg

Using this method, researchers discovered that the AI could detect whether or not two fingerprints, from different fingers, came from the same person (Wade, 2024). This system, designed by a college student, gave approximately 77% accuracy (Snodgrass, 2024), finally providing a mathematical basis for fingerprint analysis.

According to the AI analysis, the ridges at the center of each fingerprint on the same hand are similar. This allows the machine to predict a relationship (Powers, 2024).

This is not the churning out of information already in the system, as with the art and writing AIs and the accusations of plagiarism causing last year’s controversy. This provides real and necessary information that could prove useful in preventing miscarriages of justice such as the Shirley McKie case.

Conclusion

It is a shame that some people allegedly are using AI in ways that bring it into disrepute. This new technique may need more work and research, but AI provides hope for a revitalization of fingerprint analysis. This tool provides a mathematical base for fingerprint analysis. While the numbers are not yet accurate enough to permit admittance in a Court of Law, the tool has great potential for actual advancement in the science, rather than the art, of fingerprint analysis.

References

GFJC. (2013) A Simplified Guide to Fingerprint Analysis. Global Forensic and Justice Center. Sept. https://www.forensicsciencesimplified.org/prints/how.html

Mnookin, J. L. (2003) Fingerprints: Not a Gold Standard. Issues in Science and Technology 20, no. 1 (Fall 2003). https://issues.org/mnookin-fingerprints-evidence/

O’Neill, E. (2006). Mark of Innocence. The Guardian. Apr 18. https://www.theguardian.com/uk/2006/apr/18/ukcrime.features11

Powers, B. (2024) Do Prints from two different fingers belong to the same person? AI can tell. Science. Jan 12. https://www.science.org/content/article/do-prints-two-different-fingers-belong-same-person-ai-can-tell

Snodgrass, E. (2024) A College Senior used AI to try and disprove the long-held belief that all Fingerprints are unique. Business Insider. Jan 14. https://www.businessinsider.com/college-student-uses-ai-try-disprove-all-fingerprints-are-unique-2024-1?r=US&IR=T#:~:text=The%20AI%20system%20discovered%20that,belonged%20to%20the%20same%20person

Wade, G. (2024) AI Can Tell if Prints From Two Different Fingers Belong to the Same Person. New Scientist. Jan 12. https://www.newscientist.com/article/2412199-ai-can-tell-if-prints-from-two-different-fingers-belong-to-same-person/?utm_medium=social&utm_campaign=echobox&utm_source=Twitter#Echobox=1705087102

Cover Image

This includes an AI generated image found on wikicommons at https://commons.wikimedia.org/wiki/File:DALL-E_3_-_advanced_artificial_intelligence.png