would take — both for self-excluders and for those not in the program who would nonetheless be photographed. Pellizzari says that Dr. Ann Cavoukian, Ontario’s privacy commissioner, had made it very clear from the start that her office must approve the privacy component of any system installed. Thus, he made sure one of the people on his project team was from her office, and was involved every step of the way. Today, the privacy concerns have been rectified, primarily because when a gam- bler enters one of the monitored estab- lishments, his or her face is scanned by the iGWatch facial recognition system pro- vided by Oakville, Ont.-based iView Sys- tems. The image is then compared to those in the self-excluder database. If a potential match is found, security is alerted and photos of the matches pop up on their screens. It’s up to the security of- ficer to determine whether the gambler is, indeed, one of the self-excluders, and to take action. If no match is found, the image is discarded.
One of the keys to the privacy compo- nent of the new facial recognition system is software developed by Kostas Platani- otis and Dr. Karl Martin, two University of Toronto researchers and professors. Dubbed “Biometric Encryption,” it en- sures that, even in the event the data were intercepted, it cannot be tied to an indi- vidual’s personal information.
“Biometric data is more sensitive than passwords because you can’t change it,” says Martin, who is now president and CEO of Toronto’s Bionym Inc. “Once it’s com- promised, you’re compromised forever.” He explained that when a person en- rols in the self-exclusion program, the system creates a template that represents important facial features and stores it in a database along with the individual’s user ID and personal information. In a tradi- tional system, that information could be compared with templates in other data- bases to identify the individual. Biometric encryption binds a random key with the biometric data to create a unique Private Biometric Template that cannot be cross- matched, enhancing both privacy and se- curity. The key can also be used to encrypt other personal information. Only if a database match is confirmed
WWW.SECURITYMATTERSMAG.COM
will the key be released so security per- sonnel can access the rest of the indi- vidual’s personal information. For Martin, OLG’s requirements were a perfect fit for Privacy by Design (PbD), a concept developed by Cavoukian to en- sure privacy and personal control over in- formation, from initial collection through secure destruction. Privacy and security are not mutually exclusive, she says, and must be baked into applications from the very beginning.
As it relates to the OLG, the systems in- stalled were designed to increase recogni- tion rates while keeping false positives (i.e., individuals incorrectly identified as self-ex- cluders) to a minimum, to not collect un- necessary information and to protect and prevent misuse of the data it did collect. Cavoukian is satisfied with the result. “This is a technology that will offer dra- matically improved privacy protection over simple facial recognition, without compro- mising any functionality, security or per- formance — the hallmarks of a Privacy by Design application,” she insists. After two years of research, followed by integration with iView’s system, field testing and refinement, the system was pilot tested at the casino at Woodbine Racetrack (Toronto) in 2010 and full rollout began in 2011.
Although effective, the system wasn’t without challenges. Pellazzari says that lighting had to be adjusted at some lo- cations to produce adequate images,
and the entrances to some had to be tweaked to provide a “choke point” to allow everyone entering to be pho- tographed. But he points to 20 months of testing that allowed the team to iden- tify and correct these problems as a dif- ferentiator between OLG’s system and other less successful implementations of the technology.
It wasn’t the first time it had been con- sidered. In 2002, OLG had looked at fa- cial recognition systems, but rejected them over inadequate performance. It had set the match rate threshold between 65 and 80 per cent in its test environment, and the early systems couldn’t even meet those relatively modest goals. With its new system, OLG’s self-ex- cluder recognition rate is in excess of 80 per cent, and, less than three months in, with only 19 of its 27 sites live, it has al- ready intercepted almost as many people as it previously did in a full year. Imple- mentation at the remaining sites will be complete by year end. Pellizzari is happy, but he knows he can’t rest on his laurels. Audits and a third- party mystery shopper program will further challenge the system to ensure perform- ance continues to meet requirements. “We wanted to get it right,” he says. “The stakes were too high.”
And that was a gamble he wouldn’t take.
Lynn Greiner is a freelance writer in Newmarket, Ont.
FOLLOW US ON AND • SECURITY MATTERS 19
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40