Do you think you could spot the differences between a fake ID and a person standing in front of you? Now add the pressures of a lineup of impatient people, a tight schedule, and a long working day; how confident would you be in your continuous accuracy? There’s no shame in admitting you’d make mistakes.
While it’s fair to consider the negative impacts of exhaustion, poor lighting, and time constraints, the heart of the problem is this: humans are not wired to identify fraud accurately.
VERIFYING IDENTITY IS NOT AS EASY AS WE THINK
Identity verification is the bedrock of almost all of society’s critical services. Whether you’re traveling, submitting a loan application, signing up for a new service, or even applying for a job, somebody on the other side of a pane of glass or internet connection needs to confirm that you are who you claim to be.
Because identity verification has become a critical cog in the economic engine, there’s a great deal of pressure for the process to be successful, and this pressure is transferred to the shoulders of the poor souls working on the front lines. This is a problem because humans are, well, human. We get tired, we lose focus, we make mistakes, we get hungry, and whether we admit it or not, we are biased.
With the exception of a few bad apples, people like making other people happy. Even the most intimidating customs officer is susceptible to this force when a different-looking passport photo is compared to a sweet, welcoming smile. Denying a potentially life-changing loan or entrance to a country doesn’t feel nice. When those probable outcomes are faced, there’s a subconscious whisper convincing us to focus on the similarities, not the differences, between an applicant and their ID.
On top of this, forgery techniques are getting more advanced, and the list of different ID options in the U.S. just keeps growing. How can a human ever be expected to rapidly and accurately track potential signs of forgery across 28 different types of ever-changing primary IDs?
THE COST OF FRAUD IS HIGH
Cybercriminals are a clever bunch. They understand the difficulties of fraud detection and can expertly navigate through all the loopholes. This problem was already out of control pre-pandemic, but things got considerably worse when businesses struggled to adapt to the security challenges of remote work. In 2021, every $1 of fraud loss cost U.S. financial service firms $4 in damages, compared to $3.64 in 2020.
In desperation to quickly stem fraud losses, many organizations are dialing up the complexity knobs, making the conditions for verification approval much more severe. But by making verification questions more complex, you can frustrate legitimate customers and impede the growth of your business.
Stricter documentation requirements also prevent immigrants with limited paperwork from accessing services they desperately need. So not only are conventional fraud prevention strategies further increasing damage costs, but they’re also grouping innocent people in the same category as criminals.
DEFEATING BIAS WITH TECHNOLOGY
To solve the problem of conscious and subconscious bias, humans need to be removed from the verification equation and replaced with objective, logical reasoning.
AI technology meets this requirement beautifully. With AI, the legitimacy of an ID isn’t determined by color responses under different lights and the presence of key features, where conditions can be easily manipulated. Instead, every pixel of an ID document is considered to maximize the variables under analysis.
The result is an extremely low margin for error, even if an essential feature of your ID is smudged with concrete or chocolate. Additionally, machine learning allows for the instant identification of any bias as it occurs and subsequently trains algorithms to improve daily.
Because AI and machine learning provide access to a denser domain of variables, the complex feature nuances associated with aging, facial hair, and weight variance are no longer significant boundaries to accurate verification.
TRUST THE MACHINES
To appreciate the potential of this innovative approach to ID fraud prevention, consider this scenario:
While on a canoe trip, your canoe capsizes, sending your phone and wallet to the bottom of the lake. Because you can’t live without your phone, you head to the nearest store the next day to replace it. However, since you’ve lost your ID, the only proof of ownership you can provide is your phone number.
Not convinced, the representative denies your request, forcing you to order new documents and wait until they arrive in the mail before reapplying for a replacement phone. From beginning to end, the process of ID verification takes about one month.
But if an ID verification system powered by machine learning is in place, the representative can instead take a photo of your face, process it against the verification system’s database, and verify your identity in just seconds. You then return to camp, summarizing your canoe adventure on Twitter.
Until predictability and human empathy are entirely disengaged from the ID verification process, fraudsters will always find an exploitable vulnerability. Machine learning offers a means of solving the problem of fraud without abusing new business opportunities or violating human rights.
Blair Cohen is an identity-proofing expert, serial technology entrepreneur, and Founder & President of AuthenticID.