Use AI and Facial Recognition for Identity Verification & Much More
The Old Way
Starting in the Archaic days of certifying someones’ character based on family name and country origin, we have luckily evolved as humans. Most of us are used to “modern” ID verification and the memory of checking into your favorite bar/pub pops into mind.
Patrons hand over our ID to the bouncer and either get a quick, “yes” and move on or we watch the bouncer fumbling with our ID under better lighting to get a better look. Are they checking for something in particular? Humans look for certain flags and we risk score in our minds based on experience.
Computers do the same things based on logic.
Although we hear Artificial Intelligence (AI) thrown around as much as blockchain in 2017, it provides a lot of the securities a human-eye would. But, what do computers do when there is an issue with something that is seen as an anomaly?
New Way In An Evolving World
Here we are in 2019, the biggest taxi provider is an App with no employees. The biggest retailer is Amazon and they started with no retail stores! Internet banks and digital assets are fast becoming a great way to send value across borders and our digital identities are a crucial part in the Smart Economy.
Companies like Microsoft and start-ups are flocking to improve the digital verification process. To the layperson, a Driver’s License seems very secure; it may have a barcode and even some cool NFC chip and transparent signatures that make it authentic. This is all in an aim to reduce the forgery of that document, but what happens when it becomes digitalized? The first responsibility is on the GateKeeper of the digital world. If an ID is faked and certified from the beginning, then a bad actor is grandfathered into many systems. Starting with the most comprehensive certification of an individual makes everyone safer and our services we use cheaper.
Our team has been utilizing the work of the talented developers at Microsoft and their development in AI really shines over the competition.
But, as you can see in the snippet above there is room to grow.
The picture is the actor from Superbad, McLovin aka Christopher Mintz-Plasse. All we did was take a sample into their portal of “McLovin” from Superbad in 2007 and compared that to an image of him currently in ~2019.
Over 12 years has passed and either he aged really well or the machine learning detecting facial patterns for aging is still very raw. The image analysis just gathered a one year in age difference between images. It showed that the sample person aged, but not close to being accurate.
Gathering information like General Age, Sex, Emotion and Accessories are components to the arsenal. This is the core of the Microsoft stack that gives developers to build logic on top of.
Identity providers will most likely put more time into the Age and Gender bucket of logic because it is more relevant for their customers which revolve around the financial services industry.
Aging in identity verification is only one of the few issues. In Arizona; Licenses expire on the 65th birthday, and until then drivers only need new photos every 12 years — making Arizona unique in how long a license can last. Arizona has defined over 65 as aging rapidly and tries to change the rules after that threshold, but what about 16 to 28? Arguably, some of the biggest changes happen here for a lot of young men and women.
Gender is also becoming increasingly more difficult to identify based on legal and social issues around what that this is defined as.
We have heard feedback from our customers and the reports do not lie.
Backgrounds and glares affect a lot of the results for on-boarding digital users. Educating customers in how to use software is crucial before even beginning the process, but also customers do not want to read a pamphlet on how to use an App.
Unfortunately, there is not a perfect solution to these issues… but there are always tools! Image glare detection can be flagged immediately in the verification process and can prompt a user to immediately re-scan.
People may have experience with digital check scanning, hopefully with positive end results. Moving your check in proper lighting and a dark background are key to the contrast of the image so the information can be grabbed by the software.
As technology evolves and services are chained together and streamlined; the ancillary tools will be added into the loop. This includes things like Video Verification, which can train logic to detect language patterns based on dialect to detect and affirm/deny probability of the users’ region on their documents.
Together, these pieces will equal a more comprehensive risk profile for both the business and the end-user.
As this technology gets cheaper, the consumer will feel relief in their everyday life because the businesses won’t pass these regulatory requirements costs on.