Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
A 2024 study found that AI can detect similarities between different fingers belonging to the same person. That challenges an old forensic assumption, but it does not show that fingerprints are identical, interchangeable, or useless for identification. The finding could help link prints when investigators do not know which finger left them. It is not proof that an arbitrary fingerprint identifies its owner with near certainty, and it does not mean a phone enrolled with one finger will accept another.
What the researchers actually tested
The study, “Unveiling intra-person fingerprint similarity via deep contrastive learning,” was published in Science Advances on January 12, 2024. The research team included scientists from Columbia University, Tufts University, and the University at Buffalo.
The key distinction is between two different comparison tasks:
Free tools Windows power users keep installed
One-click scans. No signup required.
- Same-finger matching: Does a print come from the same finger as another impression—for example, a latent print compared with a known impression of a suspect’s index finger?
- Cross-finger person-level linkage: Do prints from different fingers, such as a right index finger and a left middle finger, belong to the same person?
Conventional fingerprint identification is mainly concerned with the first task. This research explored the second. The researchers trained a deep-learning system on roughly 60,000 fingerprint images from a public U.S. government database, arranging prints into same-person and different-person pairs. University accounts report that the system reached up to 77% accuracy on a single cross-finger pair; its performance improved when it could use multiple pairs. That figure describes the study’s particular classification task, not a universal accuracy rate for identifying people from fingerprints. The University at Buffalo’s account summarizes the dataset and reported result.
#1 Best Overall
- Target Applications - Desktop PC security, Mobile PCs, Custom applications
- Indoor, home and office use
- Blue LED - soft, cool blue glow fits into any environment; doesn't compete in low light environments
- Small form factor - conserves valuable desk space
- Rugged construction - high-quality metal casing weighted to resist unintentional movement
How can different fingerprints be distinctive and still share features?
“Unique” is often used as though it means that no two prints—or no two prints from one person—can share any meaningful structure. That is too blunt. A fingerprint can be sufficiently distinctive for comparing one finger with impressions of that same finger while also sharing broad biological characteristics with the person’s other fingers.
The model found a useful signal in overall ridge orientation and curvature, especially near the center of a print. Those broad patterns can be related across a person’s fingers even though the detailed ridge paths differ. In other words, the prints need not be duplicates for their similarities to carry information about whether they came from the same individual.
Traditional fingerprint analysis commonly emphasizes minutiae: small ridge details such as endings and bifurcations. The paper found that minutiae were almost nonpredictive for the specific cross-finger similarity task it studied. That does not make minutiae obsolete or useless in ordinary same-finger comparisons. It means the AI found a different kind of signal for a different question. The paper describes the approach and its findings.
What the “99.99% confidence” claim means
The paper reports more than 99.99% confidence in the statistical evidence that same-person fingerprints contain detectable cross-finger similarities. That is not the same as saying the system identifies any individual correctly 99.99% of the time.
Keep three kinds of numbers separate:
- Statistical confidence concerns whether the observed pattern is likely to reflect a real relationship in the analyzed data rather than random variation.
- Classification accuracy describes how often the model correctly classifies pairs under a particular test setup. The reported figure of up to 77% applies to a single cross-finger pair in the study.
- Forensic error rates would need to describe how often a validated system makes false associations or misses real ones under defined operational conditions, such as particular databases, sensors, and print quality.
So “99.99% confidence” is not a 0.01% false-match rate, a guarantee about a specific suspect, or courtroom proof beyond reasonable doubt. Nor does 77% mean the model will achieve that result on every population, sensor, or crime-scene print.
Which forensic assumption is being challenged?
Fingerprint practice has long treated a person’s different fingers as separate sources: a print from one finger is not ordinarily matched to a known impression from another finger. That is a practical and useful distinction for conventional identification, but it also left a different question underexplored: might broad features let investigators link prints to the same person even when the source fingers differ?
Rank #2
- FIDO U2F certified, and FIDO2 WebAuthn compatible for expanded authentication options, including strong single-factor (passwordless), dual, multi-factor, and Tap-and-Go support across major browsers (for services leveraging the older FIDO U2F standard, instead of using biometric authentication, Tap-and-Go allows the user to simply place their finger on the VeriMark Desktop Fingerprint Key to enable a security token experience).
- Windows Hello certified (includes Windows Hello for Business) for seamless integration. Also compatible with additional Microsoft services including Office365, Microsoft Entra ID, Outlook, and many more. Windows ARM-based computers are currently not supported. Please check back for future updates on compatibility
- Encrypted end-to-end security with Match-in-Sensor Fingerprint Technology combines superior biometric performance and 360° readability with anti-spoofing technology. Exceeds industry standards for false rejection rate (FRR 2%) and false acceptance rate (FAR 0.001%).
- Long (3.9 ft./1.2m) USB Cable provides the flexibility to be placed virtually anywhere on or near the desktop.
- Can be used to support cybersecurity measures consistent with (but not limited to) such privacy laws and regulations as GDPR, BIPA, and CCPA. Ready for use in U.S. Federal Government institutions and organizations.
The study suggests the answer can be yes. AI was able to identify a cross-finger relationship that standard approaches focused on minutiae were not designed to find. This challenges the idea that different fingers from one person are effectively unrelated for all analytical purposes. It does not establish that the entire forensic identification framework has collapsed, or that a print from one finger can simply be substituted for another.
Columbia Engineering reported that the manuscript initially met skepticism because the prevailing belief was that fingerprints were unique and cross-finger similarities could not be meaningful. That history is an account from the university, not independent proof of a field-wide failure. The published paper’s more important contribution is the evidence and method it presents, rather than the drama of the review process. Columbia’s study announcement gives its account of the work.
How it could help investigations—and what it would not establish
If validated for operational use, cross-finger linkage could give investigators another way to search for connections. For example, it might help link partial prints found at separate scenes when the exact source finger is unknown, or narrow a candidate list before detailed comparison. A person-centric search could be useful where a conventional system’s finger-specific approach misses a possible connection.
The researchers also simulated a criminal-justice lead-generation workflow and reported efficiency gains exceeding an order of magnitude in some configurations. That is a result from a simulated workflow, not evidence that police agencies have deployed the model or that it has solved real cases. A model-generated association should be treated as a lead for further investigation, not as a final identification on its own.
Real-world performance would depend on the quality and completeness of the prints, the database being searched, the sensors and image-processing methods involved, and whether the target population resembles the data used to train and test the model. A smudged or partial latent print is not equivalent to a clean recorded impression. Different capture conditions, injuries, scarring, age-related changes, or underrepresented populations may also affect performance and require specific testing.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Does this change phone fingerprint security?
Not directly. A phone that has enrolled one finger is generally designed to compare a scan with that registered finger’s template. The study’s finding that different fingers can share broad structural signals does not show that another finger will pass that stricter authentication check.
Rank #3
- 📱 QR CODE SETUP GUIDE: Scan the QR code on the packaging to access the setup page with Windows drivers and installation instructions. The package includes the main item and a Japanese manual. On the website, tap the 🌐 World icon to switch to English, then scroll down to download the English manual.
- 🚀 INSTANT ACCESS: Login 10x faster than typing passwords - Under 1 second!
- 🛡️ HIGH-LEVEL SECURITY: Match-On-Chip technology = Your fingerprint NEVER leaves the device
- 🎯 WORKS EVERY TIME: 99.999% accuracy with 360° recognition - Touch from any angle!
- 💻 PLUG & PLAY MAGIC: Zero software installation - Works instantly with Windows 10/11 Hello
The paper discusses possible future uses such as verifying a person with another finger when the enrolled one is covered, dirty, or damaged. That would be a different biometric capability and would need its own security testing. Accepting more fingers could improve convenience or resilience when one finger is unavailable, but it could also expand the ways an attacker might attempt access. The study does not establish that current phones, laptops, or payment systems use this technique.
What still needs to be tested
The results are evidence of a real cross-finger signal in the studied data, not a census of fingerprints worldwide. The dataset’s composition limits how broadly its results can be generalized. The researchers examined demographic performance and reported broadly consistent behavior across the gender and racial categories they analyzed, while also noting better performance when training and testing within the same demographic subset. That makes broader, representative validation important before relying on the method in consequential settings.
Further evaluation would need to establish performance on varied populations, sensors, image-processing pipelines, and realistic latent prints—including poor-quality, partial, and distorted impressions. Independent replication, calibrated error rates, transparent operating thresholds, audit trails, and human review would matter before an AI association could responsibly influence an investigation or a court case.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsThere are privacy implications too. A capability to infer that prints from separate contexts belong to the same person could make it easier to connect biometric records across databases. Unlike a password, a fingerprint cannot simply be changed after exposure. Any operational system would therefore need careful limits on access, retention, sharing, and how a match is used.
What the headline gets wrong
The claim that AI “proves fingerprints are not unique” turns a specific discovery into a broader conclusion the study did not establish. The researchers did not show that arbitrary people commonly share identical fingerprints, that fingerprints can no longer identify a particular finger, or that forensic evidence is invalid. They showed that fingerprints from different fingers of the same person can contain detectable shared structure—and that a deep-learning model can use it to link some cross-finger pairs.
That is a meaningful expansion of fingerprint analysis, not the end of fingerprint science. The study’s promise is a possible new way to generate investigative leads; its limits are just as important as its headline number.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
