Casino implements facial recognition to spot problem gamblers
Casino security personnel are facing a growing concern about how to keep self-confessed gambling addicts away from the tables and slot machines.
By Neil Sutton
By enrolling in “self-exclusion” programs, which are often endorsed by
the provincial regulators that govern Canadian casinos, problem
gamblers voluntarily submit their personal information and sit for
photos so casino staff can record them in a database and bar them from
Casinos involved in self-exclusion programs often rely on security
personnel to keep the faces fresh in their memory by reviewing the
database files, says Dan Morin, director of surveillance and security
for the Saskatchewan Indian Gaming Authority, a Saskatoon-based
organization that operates six First Nations casinos in the province.
“It’s not different from a photo book except it’s in a database,” says
Morin. “They go through the photos when they’ve got some downtime, they
become familiar with whoever is in that database, and when they see
somebody that they recognize, they query the database to make sure
that’s actually the person, then they go ask some questions.”
Morin estimated that about 80 people a year join a voluntarily
self-exclusion program at one of his casinos, and staff are trained to
handle the situation when a person comes forward.
It’s unusual for self-excluded patrons to come back to one of SIGA’s
casinos, but Morin estimates that there might be 15 or 20 incidents a
year. When that happens, and they’re noticed by security staff, they’re
asked to leave. If the person is a repeat offender, they may face
“Sometimes the patron may or may not agree, they may or may not be upset, but the ejection happens regardless.”
The problem for many casinos is that the system is not air-tight.
Self-excluded gamblers are still making their way into their old haunts
undetected by security staff.
The Ontario Lottery and Gaming Corp. (OLG) recently began to research
some options for improving self-exclusion programs in the province.
Through a connection with the Privacy Commissioner of Ontario, Anne
Cavoukian, they were put in touch with two researchers at the
University of Toronto who were working on breakthrough biometric
By enhancing facial recognition technology with encryption, the
researchers discovered it might be possible to not only recognize
self-excluders using software, but protect their identity by encrypting
The researchers were able to take a picture of a person’s face,
extrapolate their features, reduce them to a mathematical algorithm,
then apply an encryption key such that there’s no plain text
relationship between the person’s biometric template and their personal
“If you are self-excluded and we recognize you, then we use your face
to unbind the key to get at the meta data,” says Tom Marinelli, Chief
Information Officer for the OLG. By running through that process, you
can then put a face with the name.
Marinelli has been working with the researchers for several months and
took on private sector partners to help complete the project. Much of
the facial recognition information is being provided by German company
Cognitec, which specializes in that type of software. Canadian firm
iView Systems, which supplies database software to the casino industry,
is working as an integrator on the project as well.
The involvement of so many partners has raised the issue of who would
own the intellectual property in the event there’s a marketable
software, says Marinelli. For now, all parties have agreed to shelve
the discussion of ownership until there’s better proof of concept.
The software is currently in pilot mode, and at press time was still
being tested at OLG’s offices in Sault Ste. Marie, Ont. About 50 OLG
employees volunteered to be part the pilot and agreed to have their
faces photographed and used as test cases. Using cameras set up at the
office entrance, the test software is put through its paces and
measured for its ability to recognize faces by simply viewing them. The
number of false positives (wrong person identified) and false negatives
(right person slips by undetected) is recorded to establish a baseline
for the software’s accuracy.
The next step in the pilot process is to use the software in an actual
casino, where viewing faces is more challenging due to less than ideal
lighting. A charity casino in the Sault Ste. Marie area has been chosen
as the test environment. OLG employees on the test list will walk
through the doors to see if the software is still able to recognize
them. The test list will double in size as more employees are added —
probably staff that are already working in the casino.
The casino environment will be the ultimate test for the software, says
Marinelli. He expects it to be tested for several more months before it
can be declared a success.
“I haven’t seen anything right now that says there’s a blocker, but we
haven’t run it in a production environment yet, and everything I read
and everyone I talk to says, ‘This is a very bold thing you’re doing.’”
Casinos in other provinces have considered using some sort of facial
recognition to bolster their self-exclusion program, but there are
concerns as to its effectiveness and reliability.
“One of the perceived roadblocks in Alberta is that people are unsure
as to the reliability of the technology,” says Christian Stenner,
director of security and technology at the Silver Dollar casino in
Calgary. “People think that it still has a bit of a way to go before
it’s reliable for a self-exclusion program.”
Two years ago, the Alberta Gaming and Liquor Commission (AGLC) reviewed
its own self-exclusion program. In a subsequent report, the AGLC noted
that the effectiveness of biometric tools in identifying self-excluders
“is suspect or may not be publicly acceptable. However, the development
of other biometric technologies (e.g. retinal scans, fingerprints,
voice recognition) and growing public acceptance of tighter
identification controls may make these systems more viable in the
Dan Morin at the SLGC has come to much the same conclusion. “We did
take a look at (facial recognition) at one point, but that was probably
three or four years ago when the technology wasn’t ready yet. We use
just a manual system.”
Marinelli is familiar with this attitude but he was also a skeptic at
the start of the project. “I would have thought that we would have been
stopped at the end of the research phase.”
He also acknowledges that facial recognition may be more suitable to
casinos in Ontario than in other provinces because they are designed
such that patrons are funneled through narrow entrances. This type of
design gives security personnel the opportunity to view everyone as
they enter and could also be an ideal point to double-check faces using
cameras and biometric software.
Right now, he’s cautiously optimistic the software will prove useful,
at least in Ontario, and plans to test it in several other casinos in
the province before taking it to the next stage. “Each site’s going to
bring it’s own dynamic – it’s own lighting, where the cameras are.
We’re going to go slow to start with. After Sault Ste. Marie, we’ll
probably pick another small site.”
Earlier this year, Marinelli presented his initial findings at a
conference organized by the privacy commissioner of Ontario to
celebrate privacy breakthroughs. He hopes to be invited back.
“I’d love to come back next year and say, ‘Hey, we’ve done this.’”