Wrist motion eyed as means to detect forgeries
Software that verifies signatures via smartwatches developed by Tel Aviv and Ben-Gurion university researchers
Shoshanna Solomon was The Times of Israel's Startups and Business reporter
The handwritten signature is still the most widely accepted biometric used to verify a person’s identity. Banks, corporations, and government bodies rely on the human eye and digital devices such as tablets or smart pens to capture, analyze, and verify people’s autographs.
New software developed by researchers at Tel Aviv University and Ben-Gurion University of the Negev now enables wearable devices, currently said to be worn by one in five online adults in the US, to verify handwritten signatures, by detecting the motion of the wrist.
The study was recently published on arXiv, a site where physicists, mathematicians and other researchers post articles.
“A popular device worn by so many people should feature additional, critically useful functions,” said study co-author Dr. Erez Shmueli of TAU’s Department of Industrial Engineering. “Considering how dependent we are on signatures, we decided to develop software that would verify the smartwatch device wearer’s handwritten signature.”
Signing on a digital pad or using a special electronic pen has replaced pen and paper in many instances, but these alternatives often require cumbersome dedicated devices, the researchers say. The new software developed by Shmueli and his student Alona Levy, in collaboration with Prof. Yuval Elovici of BGU’s Department of Software and Information Systems Engineering and his student Ben Nassi, would turn any generic smartwatch into an expert signature verifier.
The new technology uses motion data to study the writer’s wrist movements through the measurement of a gyroscope — to uniquely identify them during the signing process and determine if the signature is genuine or forged.
“Using a wrist-worn device such as a smartwatch or a fitness tracker bears obvious advantages over other wearable devices, since it measures the gestures of the entire wrist rather than a single finger or an arm,” said Shmueli. “While several other recent studies have examined the option of using motion data to identify users, this is its first application to verify handwritten signatures — still a requirement at the bank, the post office, your human resources department, etc.”
When you sign, your wrist movements are unique
The team tested its system on 66 TAU undergraduates. The students, all wearing smartwatches, were asked to provide 15 signature samples on a tablet, using the tablet’s digital pen. The students were then shown video recordings of people signing during the first phase, and were asked to forge five of those signatures. They were given ample time to practice and were rewarded with bonus points for their course grade for executing excellent forgeries.
An analysis of the data showed that the smartwatch, equipped with the new verification software, was able to detect their forgeries with a “high degree of accuracy” the researchers said.
“When you sign something, you use the movements of your wrist that you have developed over years,” Shmueli said in a phone interview. “Our software is able to recognize this movement and distinguish it from those of others.”
The signing pattern is unique and this uniqueness can be captured using the motion sensors of a hand-worn device, the researchers said.
How does the technology work, given the fact that most people write with their right hand but wear their smartwatch on their left?
That is indeed an issue, acknowledged Shmueli. The software works with all wrist-worn devices, like sports devices and fitness trackers, that are usually worn on the dominant hand, he said. However, if the device in use is a smartwatch, then the user could switch the device from left to right at the time of signing. The signer could also put on a smartwatch provided by the bank or other relevant institution, Shmueli said.
“Next we plan to compare our approach with existing state-of-the-art methods for offline and online signature verification,” said Shmueli. “We would also like to investigate the option of combining data extracted from the wearable device with data collected from a tablet device to achieve even higher verification accuracy.”
The researchers have applied for a patent in an initial step toward commercializing their system.
