The Myth of Uniqueness: AI-based Study Debunks Fingerprints' Individuality
New AI-based research challenges the commonly accepted belief that fingerprints are unique, suggesting that they may not be as distinctive as previously thought
Get the latest news on fascinating discoveries, scientific advancements, and more by signing up for CNN's Wonder Theory science newsletter. Join us as we explore the universe together.
During the Covid-19 lockdowns, while waiting to start his freshman year at Columbia University, Gabe Guo was asked a question by his professor that would shape the next three years of his life. Now an undergraduate senior in Columbia's department of computer science, Guo led a team that conducted a study questioning the long-accepted truth about fingerprints: They are not all unique, as argued in a paper published this week in the journal Science Advances, with one of Guo's professors as a coauthor.
The team faced multiple rejections from journals before successfully appealing for publication in Science Advances. Despite initial pushback from the forensics community, Guo, who had no prior background in forensics, and the team continued to improve their study by adding more data, increasing accuracy, and eventually presenting incontrovertible evidence.
Revisiting Vintage Prints
Using a deep contrastive network, commonly utilized for facial recognition, the team achieved unexpected findings. They introduced their own modifications to the model and trained it using a US government database containing 60,000 fingerprint pairs, some from the same individual (but from different fingers) and others from different individuals.
In an aerial view, burned cars and homes are seen in a neighborhood destroyed by a wildfire on August 18, 2023 in Lahaina, Hawaii.
Justin Sullivan/Getty Images
How forensic investigators use DNA, fingerprints and other methods to identify remains after a fire
The AI-based system successfully identified strong similarities between fingerprints from different fingers of the same person, allowing it to accurately determine whether the prints belonged to the same individual. The accuracy for a single pair reached 77%, challenging the belief that each fingerprint is truly unique. Guo explained that the angles and curvatures at the center of the fingerprint provided a rigorous explanation for this phenomenon.
For centuries, forensic analysts have been examining characteristics known as "minutiae" in fingerprint ridges, such as branchings and endpoints, which have traditionally been used as markers for identifying fingerprints. However, while these are effective for matching fingerprints, they are not reliable for establishing correlations among fingerprints from the same individual, according to Guo. "That's the insight we've gained."
The authors acknowledge potential biases in the data and note that while they believe the AI system operates similarly across genders and races, further validation through the analysis of a larger and more diverse fingerprint database is necessary for the system to be practical in actual forensic work.
Guo expressed confidence that the discovery could enhance criminal investigations.
"This could quickly lead to new leads for unsolved cases, especially when the fingerprints found at the crime scene do not match those on file," he explained. "But it's not just about catching more criminals. It will also benefit innocent individuals who may no longer need to be subjected to unnecessary scrutiny. This, I believe, is a positive outcome for society."
Is it all just a storm in a teacup?
According to Christophe Champod, a professor of forensic science at the School of Criminal Justice of the University of Lausanne in Switzerland, the use of deep learning techniques on fingerprint images is an intriguing subject. However, Champod, who was not part of the research, expressed doubt that the study has revealed anything groundbreaking.
He said, "Their claim that these shapes are somewhat related across fingers has been recognized since the early days of fingerprinting, when it was a manual process, and it has been extensively documented for years." He added, "I believe they have exaggerated the significance of their paper due to a lack of understanding. I am pleased that they have rediscovered something known, but ultimately, it is much ado about nothing."
In rebuttal, Guo stated that no one had ever quantified or utilized the similarities between fingerprints from different fingers of the same individual to the extent that the new study has.
Guo stated, "We are the pioneers in highlighting that the resemblance is a result of the ridge orientation located in the middle of the fingerprint. Additionally, we are the trailblazers in endeavoring to compare fingerprints from distinct fingers of the same individual, using an automated system."
The system used in the study to identify similarities among fingerprints could be useful in crime scene analysis, the authors said.
Gabe Guo/Columbia Engineering
Simon Cole, a professor in the criminology, law, and society department at the University of California, Irvine, concurred that the paper is intriguing but pointed out that its practical value is exaggerated. Cole was not part of the study.
"We didn't make a mistake regarding fingerprints," he emphasized as a forensic expert. "The unverified yet seemingly true assertion that no two fingerprints are precisely identical remains unchallenged despite the discovery that fingerprints bear resemblance. It has always been recognized that fingerprints from different individuals, as well as from the same person, bear similarities."
The article mentioned that the system could be beneficial in crime scenes where the fingerprints found are from different fingers than those in the police record. However, Cole expressed skepticism, stating that this occurrence is rare since all 10 fingers and often palms are routinely recorded when prints are taken. He questioned when law enforcement would have only some, but not all, of an individual's fingerprints on record.
The team conducting the study is confident in the results and has made the AI code open-source for review, a decision that both Champod and Cole commended. Guo added that the importance of the study extends beyond fingerprints.
"It's not only about forensics, it's about AI. For as long as humans have existed, we've been examining fingerprints, yet no one ever recognized this similarity until our AI analyzed it. This demonstrates the AI's ability to automatically identify and extract important features," he stated.
"I believe this research is just the beginning of a long series of discoveries. We'll witness people using AI to uncover things that were practically invisible to us, right in front of our eyes, like our own fingers."
