Here’s a challenge for folks with a gift for remembering faces: Stand at a casino entrance and see whether any passer-by matches one of 150 mug shots of people who shouldn’t be allowed inside.
Former security executive Malcom Rutherford remembers when he thought his teams were pretty good at that task, stopping most of the people on exclusion lists from entering casinos where he worked in England and Morocco. The development of facial-recognition technology proved him wrong.
“The results are unequivocal,” said Rutherford, now executive vice president of strategic operations for eConnect, which provides artificial-intelligence, data, and video analytics to the gaming and hospitality industries. “The (facial-recognition) systems are far, far better.”
He said operators that switch to facial-recognition software identify an average of 10 times as many exclusion-list violators as they did when relying on humans to do the job. “The increase in capability from the people I’ve spoken to, based on their own data, is between 800 and 3,000 percent.”
eConnect, headquartered in Las Vegas, was founded in 2009 with an emphasis on fraud and loss prevention. It initially focused on a system that integrates point-of-sale data with digital video. Rutherford said the company’s POS Connect now operates in 450 to 500 entertainment locations. The more recently developed facial recognition software is being used or in the final installation phases at 20 to 25 sites.
Casinos face growing numbers of people on exclusion lists, some for criminal deeds, others who bar themselves voluntarily because of a gambling problem. For example, New Jersey and Pennsylvania’s Gaming Boards have each banned more than 800 people from their state’s casinos for violations. Almost all states with legalized gaming have a voluntary self-exclusion list, which can include thousands of names. Operators can face significant fines for allowing members of either type of list onto the gaming floor.
Facial recognition can be used in non-security tasks as well: allowing employees to clock in or out simply by looking at a camera, for example, or recognizing high-value players as they enter a property and alerting a host to greet them. Rutherford, who headed corporate investigations for Galaxy Entertainment before joining eConnect, calls the overall approach “identity management.”
Rutherford said a facial-recognition system also can help operators get a jump on suspicious TITO transactions. Even with a maximum TITO payout set at $2,000 per ticket, someone needs only five tickets for a $10,000 transaction that should be reported under AML regulations. Repeated instances could raise a red flag.
“This gives you a way to identify an individual without knowing their identification, kind of a facial fingerprint,” Rutherford said. “We know that over the past week, an individual cashed out an average of $10,000 a day. That basis for suspicion allows us to flag him in our facial-recognition system. The next time he arrives, we can send somebody down to talk to him.”
The company does not provide images for its system. Rutherford said operators can gather photos from local law-enforcement agencies, state regulators, or their own sources.
“We certainly give (operators) the capability to differentiate all of these groupings of individuals. Self-excluded people are different from those who are involuntarily excluded for cheating and are different from those involuntarily excluded for fighting in a bar,” he said. Although eConnect currently does not offer its own bank of information, the company is developing the ability for multiple sites within a state or casino group to share their own information seamlessly among themselves, he added.
Rutherford said current facial-recognition technology has essentially eliminated the early problem of misidentifications among racial minorities. He explained that some photo databases used to define the concept of a face for the first versions of AI software were overloaded with white men, an issue researchers understood at the time. Current databases are much more diverse and far more accurate for all. Recalling his experience at Macau Galaxy in 2012-13, he said a facial-recognition database prepared for a Chinese supplier had a similar problem: It recognized Asians far better than it did Westerners.
eConnect says its facial recogware is 99.8 percent accurate overall and accounts for differences across demographics, including race, gender, and age. In addition, it measures 42 points above the nose, so it can match faces even when someone is wearing a facemask.
