A team of roboticists asserts that, with the aid of artificial intelligence (AI), computers can determine whether fingerprints produced by different digits, such as a thumb and an index finger, belong to the same person or different individuals. This technological advancement unveils an unexpected similarity in the fingerprints of an individual, challenging conventional beliefs within the realm of forensic investigations.
The longstanding practice of utilizing fingerprints for identification, known as dermatoglyphic, relies on the assumption that the minute ridges on a person’s fingertips create enduring and distinctive patterns unique to that individual. The study delves into the question of whether prints on different fingers of the same person exhibit similarities, a notion that might seem counterintuitive given the diversity required for functions like unlocking a smartphone.
To thoroughly investigate this intriguing possibility, Hod Lipson, a prominent roboticist at Columbia University, and his colleagues harnessed the power of AI. They meticulously trained a neural network, a sophisticated program mimicking the circuitry of the human brain, first on half a million synthetic fingerprint images and then on 53,315 real fingerprints from 927 individuals. The program was intelligently designed to recognize patterns, distinguishing between an “anchor” print from one individual, a “positive” print from a different finger of the same individual, and a “negative” print from an entirely different person.
The results were quite remarkable, as the researchers observed that the AI system could identify prints from the same person 77% of the time, a performance that significantly surpassed the expected 50% success rate derived from random chance. This success underscores the potential of AI in the field of forensic fingerprint analysis.
Hod Lipson emphasizes the potential forensic application of the system, suggesting that it could be instrumental in linking fingerprint sets from different crime scenes to the same individual, even when the fingers involved are not identical. This proposition is particularly noteworthy, as it challenges preconceived notions about the uniqueness of prints from different fingers.
Moreover, the study reveals surprising similarities in the ridge orientation at the center of a person’s fingers, a ground-breaking finding that adds a new layer to our understanding of fingerprint patterns. The novelty of this discovery highlights the ability of AI not only to confirm existing knowledge but also to unveil previously unknown scientific facts.
While the results showcase the tremendous potential of AI in unravelling scientific mysteries, some experts, like computer scientist Anil Jain from Michigan State University, argue that the concept of similar prints from different fingers of the same person is not entirely ground-breaking. Jain posits that due to the influence of DNA on fingerprints, it is reasonable to expect a higher likelihood of resemblance within an individual’s prints compared to those of different individuals.
However, despite differing opinions on the ground-breaking nature of the findings, critics, including sociologist and historian Simon Cole from the University of California, Irvine, express reservations about the practicality of the AI system. Cole, in particular, raises scepticism about its application in latent print analysis, especially in cases where prints are obtained from various crime scenes. He contends that, as of now, there is no fool proof method to conclusively determine whether two prints come from the same skin, particularly considering the intricate complexities involved in latent print analysis.
In conclusion, the intersection of artificial intelligence and forensic fingerprint analysis has opened up new possibilities and challenges conventional assumptions. The study led by Hod Lipson and his team showcases the potential of AI in transforming the field, providing insights that may reshape our understanding of fingerprint patterns and their forensic applications.
Image Credit: free-images.com