ID Verification: How Accurate is Facial Recognition?

Facial recognition isn’t just a tool for tagging people on social media. It’s a technology that can enable financial institutions, law enforcement, and other government agencies to detect fraudsters, and ideally prevent them from committing fraud in the first place. Many companies have also developed solutions that harness facial recognition technology for ID verification purposes. Thanks to its high levels of accuracy, facial recognition-based ID verification has become part of many institutions’ larger KYC and AML measures.

What is facial recognition?

Facial recognition technology is able to identify faces through an algorithmic assessment of a person’s facial data. Specifically, it is able to analyze the geometry and features of the human face, connecting it to an individual identity. The most advanced facial recognition capabilities use deep learning techniques by harnessing neural networks. The neural networks “learn” from potentially millions of examples the kinds of data they are tasked to decode and draw conclusions from. As a result, the AI model learns how to discern patterns from the data points it encounters. In the specific case of facial recognition, particular facial features are isolated and then measured in relation to others to assess things like the distance between facial features, the texture of the skin, and even the thermal temperature of different areas of the face. When all these different aspects of the human face are combined, the identity of a person can be revealed. In this way, facial recognition is much like fingerprints. Just as everyone has a unique fingerprint, everyone has a unique facial composition by which they can be identified.

What is the purpose of facial recognition in ID verification?

In general, facial recognition technology can be used for two broad purposes: identification or verification. Identification is used when an unknown face is compared to a database of previously identified faces to discover the unknown face’s identity. This is frequently the kind of facial recognition that is used by law enforcement officials to identify suspects from footage from surveillance videos and the like. It generally has a lower level of precision since the inputted image or video is not always clear. On the other hand, verification is used to determine whether a person truly is who they claim to be. It is frequently encountered when a person is trying to unlock their smartphone, log in to a bank’s phone app, or confirm identity at an airport. With identity verification, the person is generally aware that their identity is being checked as part of routine measures. They have already sent a sample image of their face, so when a new action is being performed the new image is simply compared to the existing image. The accuracy here is generally higher here because the person is willing to give the camera a clear and accurate view of themselves.

How does facial recognition work in ID verification?

Here, our primary focus is on the verification side of facial recognition. That is, the following describes the steps for facial recognition analysis in the context of ID verification, where an individual consents to having his or her newly taken photo compared to an official photo. Specifically, we are speaking about ID verification for business purposes, such as opening a bank account, applying for a loan, or transferring funds.

1. A person’s photograph is taken with an emphasis on clarity.
2. The facial recognition technology analyzes the geometry of the face, including distances between different features, distinguishing marks, and skin texture.
3. A facial signature, which is essentially a mathematical algorithm, is created.
4. That algorithm is compared to a wider database of previously identified faces. According to a Georgetown University study, as of 2016, half of Americans have their faces stored in official databases.
5. A match may be made between the new photo captured and database images.

While this may be useful for determining whether a customer has criminal ties, there are now additional solutions for facial recognition in ID verification that are better at detecting identity fraud. Lightico’s ID verification solution is an example of this new method of photo ID verification. Here is how it works:

1. A call center representative sends a customer a text message containing a link that opens up to a secure mobile session.
2. The customer snaps a selfie in live mode.
3. The customer takes a picture of an official piece of photo ID, such as a driver’s license, state ID, or passport.
4. The two images are compared. If there is a match, the customer’s identity is confirmed.

The advantage of this method is that it’s far easier for both the business and customer to use (as opposed to manual checks), and doesn’t rely only on the preexisting database. It’s enough that there is a disparity between the person’s selfie and photo ID to raise red flags.

Is facial recognition accurate for ID verification purposes?

Of course, it’s not only important that facial recognition technology be easy and intuitive to use. Without high levels of accuracy, bad actors could still find their way in, and innocent people could be falsely accused of criminal intent. Fortunately, the levels of accuracy for facial recognition technology are typically extremely high. As long as there is high-quality lighting in the environment the photo is taken in, and the quality of the camera is decent, there should be no issue. Furthermore, when it comes to mobile-optimized photo ID verification, customers can always retake their photos if there are any clarity issues. According to the Center for Strategic and International Studies (CSIS), verification algorithms such as these have accuracy scores as high as 99.97%, making them just as accurate as iris scanners. This means that even financial institutions that have extremely high stakes can use it for identification purposes.

The bottom line

Facial recognition technology has very high accuracy rates, making it a suitable choice for financial institutions conducting ID verification. In addition, the latest applications of photo ID verification are easy to use for representatives and customers, ensuring that widespread adoption will become inevitable. Lightico’s photo ID verification solution harnesses the latest AI technology to allow financial institutions to verify ID through customers’ mobile phones. It’s just one of Lightico’s many customer transaction technologies including eSignatures, eForms, and digital payments. Learn more at Lightico.com.