Top Biometrics Trends and How They Approach User Privacy

Data protection, in combination with biometrics, has become a powerful cocktail – for both the right and wrong reasons.

On the one hand, the unprecedented capabilities of biometric identification could provide a big reason for today’s businesses to smile. However, many of them sweat hard trying to prevent legal issues, such as lawsuits based on the Illinois privacy protection act. Having one’s face saved in a database for unknown purposes can make consumers feel uncomfortable – not necessarily because it could be misused, but because a person’s likeness feels out of control.

That’s why, in recent months, many innovations have sprung up to bring a level of control over the use of biometric data. The combination of cryptography and secure communication has helped introduce the so-called self-sovereign identity. It brings a person’s identity fully under user control, giving them the ability to choose how they want to share their ID traits.

Biometrics is a natural companion to access self-sovereign identity, as it can prove, without fail, that the owner is always identified. Now, EU countries are amongst those already discussing how to put self-sovereign identity into practice.

The key to achieving that is increasing computing power and specialized chipsets. These enablers of edge computing can offload the computing necessary for biometrics and edge devices, such as cameras. This means that the data sent over the network doesn’t encompass entire video streams but only pre-processed faces, extracted into the form of templates. These are both smaller than the actual image – drastically cutting bandwidth requirements – and cannot be used to reconstruct the original face. Even if intercepted, the data would be useless to the hacker.

Advanced computing enables the matching of extracted data on the cloud, reducing demands for own, on-site hardware and its related security frameworks. On top of that, servers hosted in large cloud depositories usually have much better default security than servers maintained by small providers. This way, businesses can delegate some security responsibilities to the cloud provider.

Last but not least, computing power and advances in neural networks allow for using data that isn’t really personal to anyone. By using neural networks, companies can generate datasets from synthetic faces and other images necessary for training neural networks (IDs, for example). Websites like This Person Does Not Exist allow for creating realistic-looking yet non-existent faces. These can replace existing training datasets or augment them.

For example, a biometric company could use such images to generate more faces of a certain skin complexion to improve their dataset and avoid bias against certain groups of people. The generator can also take care of specific cases that would otherwise be hard to obtain, such as training facial recognition on people wearing face masks, as the masks can be digitally added to images in a realistic way.

Other programs can rotate faces and show them from different angles, further improving facial recognition algorithms. Although facial datasets aren’t physically encoded within the trained algorithm, there can be problems with obtaining or rescinding consent from the person in the training dataset. Synthetic faces can solve such problems and even balance out possible biases – finally tackling one of the biggest skeletons in biometrics’ closet.