How to Choose Your Facial Biometrics Solution?

In an increasingly interconnected world where social interactions are becoming more digitized, identity fraud has become commonplace. The latest case involved an employee of a multinational company in Hong Kong who cost his company $25 million. This employee conducted several banking transactions before realizing that his supposed boss was actually an AI (a deepfake).

So how can you ensure that the person you're interacting with is indeed the right person? Among the myriad of technologies available to authenticate your counterpart, facial biometrics emerges as the most promising solution. To counter identity fraud, facial biometrics allows you to: 

• detect that the right person is truly behind their screen through liveness detection 

• verify that they are indeed who they claim to be through facial comparison

Faced with increasingly frequent and sophisticated identity theft threats, you should certainly not take lightly the choice of your facial biometrics solution to protect your users. So, what criteria should you consider to implement a solution that meets your needs? Follow the guide!

A Facial Biometrics Solution to Ensure Maximum Security for Your Users

As we mentioned earlier, digital interactions are now part of our daily lives. However, the flip side of the coin is that the digitization of services has made identity theft possible and, more importantly, easy. In 2023, several thousand attempts of identity fraud were recorded in France, resulting in several billion in losses for both businesses and individuals. Offering a proven security solution is therefore vital to protect your company and your users.

Make Precision and Reliability Your Priority

Facial biometrics has the advantage of being very precise, but like any technology, there is no such thing as zero risk. There are only thresholds that certify the reliability of a solution.

For liveness detection, we talk about:

• APCER, for Attack Presentation Classification Error Rate, which is the proportion of attacks incorrectly classified as a real person. The higher this rate, the more compromised the solution's security. This means it allows many attacks to pass.

• BPCER, for Bonafide Presentation Classification Error Rate, which is the proportion of real people incorrectly classified as attacks. The higher this rate, the worse the user experience, as it means many people are wrongly rejected.

For facial comparison, we talk about:

• FMR, for False Match Rate (or FAR, for False Acceptance Rate), which represents the probability that a pair of faces are considered to match when they are actually different people. The higher this rate, the less reliable the solution.

• FNMR, for False None Match Rate (or FRR, for False Rejection Rate), which represents the probability of rejecting a pair of faces when they belong to the same person. Like BPCER, this rate has a strong impact on the user experience.

Ideally, the levels of false acceptance or rejection should be as low as possible to ensure both maximum security and a good user experience. However, if there is a compromise between the two data, security should be prioritized first, as it is better to wrongly reject a person than to accept someone who is not who they claim to be!

Evaluation organizations attest to the reliability of facial biometrics solutions when they meet ISO 30107-03 standards, both for liveness detection (e.g., iBeta) and for facial comparison (such as NIST).

What about Personal Data Protection?

The CNIL, the French National Commission on Informatics and Liberty, defines biometrics as "the set of computer techniques allowing the automatic recognition of an individual based on their physical, biological, or even behavioral characteristics." As you can imagine, there is nothing more personal than a person's face or their identity document! These highly sensitive data benefit from special protection status, as stated in Article 9 of the General Data Protection Regulation (GDPR).

Make sure to choose a facial biometrics solution that complies with European regulations and ensures data storage in Europe!

A Facial Biometrics Solution to Enhance User Experience

Facial biometrics is a relevant solution for a variety of use cases, such as authentication or onboarding. But regardless of the case, user experience is a real concern that impacts conversion rates and customer acquisition costs.

A Universal Solution that Works Everywhere and for Everyone

Universality is a crucial criterion when it comes to facial biometrics to optimize your users' experience. Firstly, your solution must be effective for everyone, without any discrimination. An algorithm is only as intelligent as the data it's trained on. Ensure that your solution is tested on a wide range of people, of all ethnic backgrounds and ages.

Secondly, your facial biometrics solution must be able to work on any type of device. Not everyone necessarily has the latest iPhone in hand; some may not even have a smartphone and only use computers. This leads us to the next point...

...Your solution must be effective regardless of the quality of the media (photo or video), and regardless of capture conditions. Imagine having to ask your users to go out in direct sunlight just to authenticate... That's a guaranteed drop in conversion rates!

A Passive and Fast Solution

As mentioned in our previous article, the primary challenge in fighting identity theft is to ensure that the right person is physically present in front of the camera. This liveness detection system comes in two forms:

• An active form, which requires the user to perform an action such as a head movement.

• A passive form, which requires no action.

And it's this second form that optimizes your clients' experience. Remember, the less you constrain your users, the more you'll improve your conversion rates.

Find out in this article the most common facial biometrics attacks.

Another point of vigilance that plays a role in user experience is time. Between capturing the media, sending it to the server, analyzing the media, and receiving a response, you can quickly lose your users. Let's draw a parallel with the web: according to a Google study, 53% of internet users leave a page if it takes more than 3 seconds to load!

Offering a fast solution, both for media capture and analysis, will greatly reduce friction and improve your conversion rates.

Choosing a Flexible and Easy-to-Integrate Facial Biometrics Solution

Finally, the last criterion for choosing your facial biometrics solution is integration. You have two options:

• Integration where you only need the API of your facial biometrics solution to send it the necessary information for media analysis.

• Using the SDK of your solution, which allows you to capture the media in addition to the API needed for information processing.

We recommend using the SDK of your solution because this ready-to-use user interface reduces the risk of error when capturing media. Furthermore, it allows for faster integration and access to a wider range of features.

Depending on your infrastructure, you have the option to deploy your facial biometrics solution in two ways: as "SaaS" or as "On-Premise", each with its own advantages and disadvantages.

A SaaS Deployment

In SaaS (Software as a Service), the solution runs on your facial biometrics provider's cloud, which they maintain.

SaaS deployment is advantageous as it is very quick to integrate, less costly due to the minimal infrastructure to manage, and easier to deploy on a large scale. However, data necessarily transit through an external cloud. As mentioned earlier, make sure to choose a solution that complies with GDPR.

An On-Premise Deployment

An "on-premise" deployment means that the solution is installed on your own infrastructure (on your machines).

The advantage of an "on-premise" deployment is that you have control over your data and flows. Conversely, this integration takes longer to set up, is more costly (in terms of infrastructure and human resources), and is more difficult to scale.

Conclusion:

With increasingly sophisticated identity theft threats, it is imperative to offer your customers an effective, seamless, and easy-to-integrate facial biometrics solution. In this regard, the Unissey solution ticks all the boxes mentioned throughout this article.

Our teams are at your disposal to test our facial biometrics solution for free. Contact us !