Samsung patents fingerprint system that learns from fake and dirty prints
This patent describes a phone or device that stores fingerprints as digital data points rather than just pictures. When you scan your finger, the device uses an AI model to create 'virtual' fingerprints that look like your finger might appear when it is dirty, wet, or if someone tries to fake it with a mold.
In the filing’s own words
An electronic device is provided. The electronic device includes a sensor, memory configured to store plurality of fingerprint templates, each of the plurality of fingerprint templates is an embedding vector indicating each of plurality of registered fingerprint images, and at least one computer program, and at least one processor communicatively coupled to the sensor and the memory, wherein the plurality of registered fingerprint images comprises a forged fingerprint image made by forging a fingerprint of a user, an abnormal state fingerprint image comprising an obstacle obstructing fingerprint authentication, and a normal state fingerprint image which is either a successfully registered fingerprint image or a fingerprint image matching the successfully registered fingerprint image, and wherein the at least one computer program includes instructions that, when executed by the at least one processor individually or collectively, cause the electronic device to acquire, using the sensor, an input fingerprint image indicating a fingerprint image input by the user, identify whether the input fingerprint image matches at least a part of the plurality of registered fingerprint images based on the plurality of fingerprint templates, in case that the input fingerprint image matches the at least a part of the plurality of registered fingerprint images, generate at least one virtual fingerprint image, based on the input fingerprint image matching the at least a part of the plurality of registered fingerprint images, store a template of the input fingerprint image and at least one template of the at least one virtual fingerprint image in the plurality of fingerprint templates, and train a fingerprint generation artificial intelligence model to generate a fingerprint image similar to the at least one virtual fingerprint image or the input fingerprint image.
What exactly they claim (claim 1)
1. An electronic device comprising: a sensor; memory configured to store a plurality of registered fingerprint templates, each of the plurality of registered fingerprint templates comprising at least one embedding vector indicating a respective one of a plurality of registered fingerprint images, and at least one computer program; and at least one processor communicatively coupled to the sensor and the memory, wherein the plurality of registered fingerprint images comprises a forged fingerprint image made by forging a fingerprint of a user, an abnormal state fingerprint image comprising an obstacle obstructing fingerprint authentication, and a normal state fingerprint image which is either a successfully registered fingerprint image or a fingerprint image matching the successfully registered fingerprint image, and wherein the at least one computer program comprises instructions that, when executed by the at least one processor individually or collectively, cause the electronic device to: acquire, using the sensor, an input fingerprint image indicating a fingerprint image input by the user, generate an input fingerprint template from the input fingerprint image, compare the input fingerprint template and the plurality of registered fingerprint templates, identify whether the input fingerprint template matches at least a part of the plurality of registered fingerprint templates, based on the comparison, in case that the input fingerprint template matches the at least a part of the plurality of registered fingerprint templates, generate, using a fingerprint generation artificial intelligence model, at least one virtual fingerprint image representing a fingerprint of the user in at least one of an abnormal state or a forged state, based on the input fingerprint image, generate at least one virtual fingerprint template from the at least one virtual fingerprint image, store the input fingerprint template and the at least one virtual fingerprint template in the memory, and train the fingerprint generation artificial intelligence model based on the input fingerprint image.
Facts
- Company
- Samsung (Samsung Electronics Co., Ltd.)
- Official title
- Electronic device for managing enrolled fingerprints and method for the same
- Patent number
- US12738102B2
- Application number
- 19213340
- Filed
- May 20, 2025
- Granted
- Sep 15, 2026
- Inventors
- 2
- Claims
- 18
Classified under
G06V 40/50 Physics & computingG06F 21/32 Physics & computingG06V 10/82 Physics & computingG06V 40/1318 Physics & computingG06V 40/1365 Physics & computingG06V 40/40 Physics & computingMore from Samsung in Security & privacy
- Samsung patents SSDs that use AI to spot ransomware attacks Oct 20, 2026
- Samsung patents AI system to find and fix device security flaws Sep 17, 2026
- Samsung patents method to find lost devices via nearby phones Sep 8, 2026
- Samsung patents video camera that tracks authorized faces for security Sep 1, 2026
- Samsung patents system that ranks home camera clips by unusual events Sep 8, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.