91av

IS YOUR signature exactly the same as it was a decade ago? Or even last
month? Probably not. Signatures change, albeit in minor ways, and that’s a big
problem for the new breed of “biometric” electronic devices that use signatures,
rather than hard-to-remember PINs or passwords, to identify people.

But now there’s a signature recognition system that can cope with natural
changes to the way you sign your name. “Your signature varies not just over long
periods of time, like ten years, but also over short periods of time depending
on factors like your mood,” says Jony Bassan, product development manager at
WonderNet, a high-tech start-up based in Bnei Brak, Israel. “The fact that you
may be standing in an awkward position when you are signing, or simply find that
there isn’t enough room on a shop counter to sign properly, may also influence
how you sign.”

WonderNet’s new PenFlow system uses specially written software linked to
existing electronic pen hardware from the Japanese company Wacom, which is one
of WonderNet’s backers.

The movements of the electronic pen are picked up by a grid of conductors in
a flat graphics “tablet”, and sent to the PenFlow software for analysis. The
program picks out information on dozens of factors, including the pen’s stroke
rate, the rhythm of its movement and even the height the nib hovers above the
tablet between words. Each factor is stored digitally, and together they make up
a complete, unique description of an individual’s signature This is then
encrypted, so that it can be safely stored and transmitted without falling prey
to hackers. No pictorial image of a sample signature need ever be stored.

Competing manufacturers are working along similar lines, but WonderNet has a
patent pending on an additional technique that allows PenFlow to compensate for
variations in people’s signature. The software initially requires three specimen
signatures, and then gradually learns how the signature changes, by assessing
its “propensity to fluctuation”.

If someone signs at a greater angle than usual, says Bassan, the PenFlow
software checks whether it is an authentic signature that has been
“rotated”—and calculates the probability that the signature is real.

Alex Herman, operations manager at WonderNet, claims that PenFlow is a big
advance on past attempts to develop signature-analysis systems which rely on
software that recognises stored images. “Image recognition just doesn’t capture
the distinctive way in which you sign,” he says. This allows fraudsters to learn
to copy signatures well enough to fool the software.

Erik Rees, chairman of the British Institute of Graphologists, says that
PenFlow would be a first if it can indeed judge slight differences in
signatures. But he warns: “Once it has rejected a signature, I think a further
judgement of its authenticity should be made by a human being.”