Face-INT

The Need

What was once considered science fiction, has in many cases become a reality over the years: from automatically opening sliding doors to touchscreen tablets, invention sometimes leaps from the page into reality. This is also true for the field of biometrics and facial recognition technologies - on one hand, a super-advanced and convenient high-tech tool for the identification for users, but at the same time - a tool that might be abused and used for malicious purposes, from stealing sensitive data, to taking over someone's entire identity.

In the ultra-modern technological age we live in, each of us is required to identify daily, whether online (on banking, insurance, credit card, health services sites and more), or physically - at work, the gym, the airport, and sometimes even at the entrance to our own home.

In fact, on a global scale, according to estimates, this outdated and mandatory use of id cards and passwords consumes about 1,300 years of cumulative human life (!). And if that wasn’t enough, any such identification procedure is also a wide open backdoor which may be prone to phishing attempts (in which fictitious websites impersonate reputable services, and try to steal the user’s login credentials, credit data or passwords).

Despite the speed, elegance and convenience built into the very concept of face recognition technologies, which are based on a way of identification that can not be forgotten or lost, saving the need for technical support whenever a password is forgotten, these systems haven't managed yet to become a real substitute for personal identification needs, mainly because of the issue of users' privacy and related strict regulatory requirements - companies are not allowed to keep sensitive personal data, which a third party could potentially decipher.

Another market desperate to solve this inherent issue is the rapidly evolving field of IoT (Internet of Things), which includes a vast variety of "smart" devices, many of which require a user identification procedure to operate, or exit "lock" mode, where face recognition tech can efficiently replace many items such as car keys, house keys, various coders and more.

The Solution

In a truly innovative breakthrough for the field of biometric face recognition, Face-INT has developed a unique encryption algorithm, which enables  - for the first time - biometric identification without storing any data, saving the need to keep or secure the inherently sensitive biometric info, in accordance with global regulatory requirements.

The company's patented solution actually turns this biometric data, using a hash function, into a very long string of numbers, from which the original data cannot be recovered in any way.

Customers using Face-INT's technology can, following a short registration process via mobile phone, identify themselves easily and quickly, at any time, using only their face.

How does it work?

In addition, the Face-INT system can continue to identify users even when their outer appearance changes slightly, thanks to the use of advanced ML (machine learning) components, and knows how to "bridge" over changes such as growing a new beard, wearing glasses, etc., and even has the ability to overcome possible spoofing attempts.
 

Technology

To fully understand the uniqueness of Face-INT's technology, keep in mind that every face recognition system has an engine, whose function is to process each face image and output it as a set of values ??(face descriptors), which reflect and characterize the same person - mathematically speaking. For the sake of abstraction, let's say each value expresses some important feature - for example, the size of each organ, the distance between the eyes, the distance between the ears and nose, etc.

In order to identify people in the system, the same values ??must be saved initially, at the registration stage, and a mathematical function comparing the set of values ??at the entry stage must be activated, for each of the sets stored in the system - from the registration stage, until the most identical set is found.

But these face recognition engines can not accurately reproduce the same values ??for the same person every time, since any slight change in the camera conditions (such as angle, lighting, facial expressions), and even more so the changes in a person's appearance over time (hair or makeup, glasses, beard, etc) - cause the data itself to change.

Despite the different results, the function which compares the data sets produces a result of a relatively small "gap" between different images of the same person, and a high difference between images of different people. Based on this figure the system can provide the correct identification outcome.

The main problem with previous face recognition systems, as mentioned, is that the storage of those sets of values in a database, constitutes a barrier in terms of existing regulatory demands around the world, aimed at protecting user privacy.

Standard face recognition system - illustration diagram

Unlike the standard process, using Face-INT's method, the algorithm is able to produce a fixed set of values for each image of the same person, and learns to ignore different camera conditions as well as any natural changes in the person's appearance over time.
 

Face-INT's technology - illustration diagram:

This unique ability to reproduce the same fixed (deterministic) set of values, for the same person, makes it possible not to have to save any biometric data, through another step - hash encryption.

Unlike standard encryption, which is defined as two-way encryption, where you can encrypt a value and also decrypt it using the same encryption key, hash encryption is defined as a one-way encryption, which means the encrypted value can't be unlocked in any way, making this method the safest and most common method for password storage, in various systems.

The process in the stage of registration:

1. Photographing the user - The Face-INT engine produces a fixed set of values
2. The set of values is then encrypted via hash
3. The values set is stored in the database, meaning even a possible hack into the system will not provide any biometric / personal data usable in any way by hackers

After registration, each time the user logs in, the system performs the following actions:

1. Photographing the user - The Face-INT engine produces a fixed set of values again
2. The set of values is then encrypted via hash
3. The system scans the database and looks for a match for approval

Existing face recognition systems don't currently use hash this way, since Face-INT is the first to develop an algorithm that allows the creation of an encrypted set of values which has the ability to remain identical, even between the registration process and later logins. Without this unique and exclusive development, every single minimal change in the person or camera - produces a completely different encrypted value, making it unusable.

For the purpose of simple password encryption, where the password entered into the system is the same every time, this method is easy and simple to apply, but for complex face recognition where the data obtained is "noisy", other systems simply can not produce a fixed set of values, and therefore of course can not encrypt it in this method, as well.