- AI challenges unique fingerprints
- Forensic community skeptical
- Possible impact on cold cases
An advancement in artificial intelligence fingerprint analysis has the potential to resolve thousands of unresolved disputes.
A computer using an artificial intelligence system disproved decades-old conventional wisdom that every individual fingerprint is unique.
Even if a thief left a thumbprint at one crime scene and an index finger print at another, it would be impossible to establish a connection between the two.
A student at Columbia University made the groundbreaking discovery while attempting to determine whether or not artificial intelligence could establish connections between ostensibly dissimilar fingerprints belonging to the same individual.
To test the hypothesis, engineering graduate Gabe Guo, who lacked a background in forensics, displayed a computer containing pairings of images representing approximately 60,000 fingerprints.
In certain circumstances, fingerprints would be collected from two distinct digits on the same individual’s hand; in other instances, they would be collected from distinct individuals.
As time passed, the computer became capable of identifying telltale patterns in which two fingerprints with radically different appearances originated from the same hand – an unprecedented occurrence.
In a statement, Guo et al. state, “Our primary finding is that fingerprints from various fingers of the same individual exhibit striking similarities; these results hold true for all finger combinations, including those from different hands of the same individual.”
The forensics community initially expressed disapproval towards the findings.
The research was denied publication in a reputable forensics journal, where an anonymous expert reviewer and editor reached the conclusion that “since it is common knowledge that each fingerprint is unique, it is not possible to detect similarities even if they originated from the same individual.”
However, colleagues and Mr. Guo persisted.
Professor of Engineering at Columbia Hod Lipson stated, “While I don’t typically contest editorial decisions, this discovery was too critical to disregard.”
If this information alters the scales, authorities might reopen dormant cases, potentially leading to the acquittal of innocent individuals in some instances.
A significant obstacle that contributed to the denial of the research finding was the lack of clarity regarding the information utilized by the AI to establish a connection between seemingly unrelated fingerprints that had eluded forensic analysis for decades.
The group reached the conclusion that the AI had detected previously unseen patterns in the ridges of the fingerprint’s centers.
Professor Lipson characterized the research as an instance of an innovative realization resulting from artificial intelligence.
He stated, “Many individuals believe that AI is incapable of truly discovering new information; rather, it merely regurgitates it.”
However, this study demonstrates how a relatively uncomplicated artificial intelligence (AI) can yield insights that have eluded specialists for decades when provided with a relatively plain dataset that has been circulating in the research community for years.
“Even more exciting is the fact that an undergraduate with no prior experience in forensics can use artificial intelligence to successfully challenge a widely held belief in an entire field,” he continued.
“An explosion of scientific discovery led by non-experts using AI is imminent, and the expert community, including academia, must prepare.”
There is an additional prospective application for the discovery: computer fingerprint identification. Certain security systems and portable computers recognize individuals by fingerprint. If, however, the finger used to create the original print is impaired in any way—for example, by a bandage—the user is unable to access the system.
