Fujitsu Laboratories Ltd. announces the development of a facial recognition technology that uses conventional cameras to successfully identify efforts to spoof authentication systems. This includes impersonation attempts in which a person presents a printed photograph or an image from the internet to a camera.

Conventional technologies rely on expensive, dedicated devices like near-infrared cameras to identify telltale signs of forgery, or the user is required to move their face from side to side, which remains difficult to duplicate with a forgery. This leads to increased costs, however, and the need for additional user interaction slows the authentication process.

Convenience of face authentication

To tackle these challenges, Fujitsu has developed a forgery feature extraction technology that detects the subtle differences between an authentic image and a forgery, as well as a forgery judgment technology that accounts for variations in appearance due to the capture environment.

Fujitsu's new technology ultimately makes it possible to prevent impersonation with forgeries using only face images taken at the time of authentication, enhancing security without sacrificing the convenience of face authentication and contributing to the DX (digital transformation) of operations with improved personal authentication technologies.

Other authentication methods

It has proven difficult to determine the authenticity of a face by relying on a facial image alone

While biometric authentication continues to grow in popularity, many risks remain. In some cases, when facial images are disclosed on the Internet via SNS, etc., the possibility emerges that the image may become the target of malicious users if stolen due to the loss of an ID card with a facial photograph, etc.--because of the prevalence of such images, this makes facial authentication more vulnerable than other authentication methods, such as fingerprints or palm veins.

Smartphone screens, ID cards, and face images printed on paper vary in their appearance due to factors like reflections or blurring on a smartphone screen. It has proven difficult to determine the authenticity of a face by relying on a facial image alone because of the effects of similar fluctuations, such as reflections caused by fluorescent lights or sunlight, or blurring caused by facial movement.

General-Purpose cameras

For this reason, special cameras like near-infrared cameras or depth cameras that measure the distance between the subject and the camera are used to catch typical signs of forgery. These methods remain imperfect, however, and lead to issues including increased costs for dedicated cameras and reduced convenience due to the addition of motion information required when using general-purpose cameras.

The development of technologies that can conveniently and inexpensively detect spoofing with general-purpose cameras has become a topic of consideration. Fujitsu has developed a technology that can detect the impersonation of others through photographs, etc. from face images taken with a general-purpose camera. The features of the developed technology are as follows.

Forgery feature extraction

Image processing technology is used to digitize the characteristic features of forgery

Various features characteristic of a forgery remains in images obtained by presenting the forgery to the camera, such as reflections on the terminal screen of a smartphone, and distortion of the shape of the face caused by taking a planar forgery. Fujitsu has developed a forgery feature extraction technique to express the difference between the forgery's characteristic features and the real face as determinable values.

First, the face image captured by the camera is separated into various elements that exhibit the characteristic features of forgery, such as reflection elements and shape elements. Next, image processing technology is used to digitize the characteristic features of forgery for each of the separated elements, and the characteristics of each element are combined to generate a characteristic for judgment. This makes it possible to identify counterfeits without information based on user operations.

Deep learning techniques

In the past, in order to respond to variations in image appearance caused by the capture environment, a single determination model was generated by training a system with face images containing various variations using machine learning. However, the wide range of variations in the way images are taken, depending on the type of forgery, such as a smartphone screen or ID card, complicates the boundary between the real face and the forgery, making it difficult to identify the forgery even with the latest Deep Learning techniques.

Therefore, Fujitsu has developed a technology that can correctly identify counterfeits by generating determination models that reduce the influence of variations by learning the categories of face images that have similar variations, such as face images taken at the office or face images taken by a window.

Counterfeit with machine learning

The development technology steps are divided into a training phase and a judgment phase. In the training phase, face images acquired in various environments are classified into categories such as window, backlight, and normally based on the capture environment, such as the intensity of light and the direction of light.

Next, a judgment model is made for determining whether the target is a real face or a counterfeit with machine learning, using the decision features generated by the forgery feature extraction technology for each category.

Inexpensive spoofing detection

The similarity between the input image and each category is calculated dynamically

In the judgment phase, in order to estimate which of the categories defined in the training phase the input image capture environment is close to, the similarity between the input image and each category is calculated dynamically. Next, in order to emphasize the result of the determination model of the category in which the input image and the environment are close to each other, a value obtained by multiplying the score indicating the authenticity output from each determination model by the weight based on the similarity with each category is used to determine whether or not the object is a fake.

By using these technologies, it becomes possible to identify counterfeits using only the information of face images taken by a general-purpose camera and to realize relatively convenient and inexpensive spoofing detection.

Personal authentication technology

Fujitsu reviewed its own evaluation data set collected in a general office environment or an environment where telework outside the office is assumed and confirmed that spoofing attempts can be detected with the same level of accuracy as before, even when there is no specified movement by a dedicated camera or a user. This makes it possible to prevent unauthorized access at a low cost without sacrificing convenience.

The technology offers the potential to improve security for workers remotely accessing company systems from offsite and to contribute to the digital transformation of operations through the advancement of enhanced personal authentication technology. Fujitsu aims to further improve the accuracy of its forgery detection technology with the aim of putting it into practical use by the end of the fiscal 2020 in March 2021.

Download PDF version Download PDF version

In case you missed it

What Is The Role Of Gunshot Detectors In The Security Ecosystem?
What Is The Role Of Gunshot Detectors In The Security Ecosystem?

Sadly, active shooter incidents have become so common that they no longer grab big headlines or dominate the news cycle. A near-constant cascade of active shooter events persists in the background of our collective consciousness, a familiar drumbeat that is no less tragic because it is continuous. As more active shooter incidents occur, the security marketplace continues to implement solutions to minimize the impact, including gunshot detection. We asked this week's Expert Panel Roundtable: What is the role of gunshot detectors in today’s security ecosystem? 

Mythic’s AI Chip Leverages Analog Technology For Faster Speed, Less Power
Mythic’s AI Chip Leverages Analog Technology For Faster Speed, Less Power

For security professionals who thought analog systems were a thing of the past, a new approach by Mythic Inc. demonstrates that everything old is new again. Using older technology in a new way, the Mythic M1076 Analog Matrix Processor leverages analog computer chips from a previous generation to drive new levels of artificial intelligence (AI) performance with lower power requirements.  Low power and high speed  Mythic provides power-efficient AI at the edge, including inside video cameras. The design combines embedded flash memory with analog computing power to achieve faster AI processing, supporting up to 25 trillion operations per second (TOPS), with the very-low power levels conducive to edge devices. The scalable, single-chip analog compute-in-memory architecture provides high-performance inference without consuming the power and energy that digital solutions require to move data at high speeds between separate processing and storage components. Single-chip design  High-resolution video analytics with low latency, comparable to a GPU, is provided by AI, but at 10 times less power “We use a different approach to processing and storage by resurrecting analog technology for faster computing power in a limited size and cost,” says Tim Vehling, Senior Vice President, Product and Business Development at Mythic. The Mythic chip solves several design challenges for camera manufacturers. The single-chip design with no DRAM (dynamic random-access memory) caters to limited space requirements. High-resolution video analytics with low latency, comparable to a graphics processing unit (GPU), is provided by AI, but at 10 times less power than a typical system on chip (SoC) or GPU. The typical 3-4-watt power draw is consistent with a limited power budget for power over Ethernet (PoE). Passive heat dissipation does not require active thermal management. Applications of the analog chip For video applications, the chip provides faster speed to accommodate more cameras, more resolution, and more details in images. In addition to providing scalability, the chip supports a variety of host platforms, including X86, NVIDIA Jetson Xavier NX/TX2, Qualcomm RB5, and NXP i.MX8M. It supports Linux Ubuntu 18.04 and Linus for Tegra (NVIDIA) operating systems. The chips can plug into NVIDIA or Qualcomm platforms to enhance AI capabilities for a variety of applications. The chip also has utility in other deployments, including drones, where Mythic works with the Qualcomm RBS platform to enable multi-thousand-dollar drones for larger applications. Integration into devices The chip can augment the capabilities of a CPU without replacing it or completely redesigning a product The chip handles image sensing, multiple cameras, radar, and lidar sensors, in addition to flight navigation, control, and communication, in addition to in-flight analytics. Inside NVRs, Mythic chips provide high-level processing at a fraction of the cost, says Vehling. Integration of the technology into cameras and other products is simple – it simply plugs into an M.2 expansion slot, and the software is downloaded to drive the AI algorithms. The chip can augment the capabilities of a CPU without replacing it or completely redesigning a product, in effect providing an instantaneous improvement in performance. No shortage Because Mythic uses older technology, there are no shortages compared to some later-generation chips. The 40-nanometer chips are a mature technology, manufactured in Japan, while newer processors are smaller at 5 or 7 nanometers. The newer chips are more likely to be in short supply. The Mythic M1076 chip is currently being evaluated but is not yet in production. The company expects to be shipping the product in the second half of 2022, and it will be sold to camera manufacturers and other OEMs to be incorporated into their products.  Adds value inside cameras For security end-users, Mythic’s AI chips will add new value inside video cameras and other equipment in terms of better performance, small size, and less power. For integrators, the technology will expand equipment options, such as providing high-level analytics in cameras while requiring only 2 to 3 watts of power, consistent with the use of PoE.

Why Face Recognition As A Credential Is The Ideal Choice For Access Control?
Why Face Recognition As A Credential Is The Ideal Choice For Access Control?

In the field of access control, face recognition has come a long way. Once considered too slow to authenticate people's identities and credentials in high traffic conditions, face recognition technology has evolved to become one of the quickest, most effective access control identity authentication solutions across all industries. Advancements in artificial intelligence and advanced neural network (ANN) technology from industry leaders like Intel have improved the accuracy and efficiency of face recognition. However, another reason the technology is gaining traction is due to the swiftly rising demand for touchless access control solutions that can help mitigate the spread of disease in public spaces. Effective for high volumes Face recognition eliminates security risks and is also virtually impossible to counterfeit Modern face recognition technology meets all the criteria for becoming the go-to solution for frictionless access control. It provides an accurate, non-invasive means of authenticating people's identities in high-traffic areas, including multi-tenant office buildings, industrial sites, and factories where multiple shifts per day are common. Typical electronic access control systems rely on people providing physical credentials, such as proximity cards, key fobs, or Bluetooth-enabled mobile phones, all of which can be misplaced, lost, or stolen. Face recognition eliminates these security risks and is also virtually impossible to counterfeit. Affordable biometric option Although there are other biometric tools available, face recognition offers significant advantages. Some technologies use hand geometry or iris scans, for example, but these options are generally slower and more expensive. This makes face recognition a natural application for day-to-day access control activities, including chronicling time and attendance for large workforces at construction sites, warehouses, and agricultural and mining operations. In addition to verifying personal credentials, face recognition can also identify whether an individual is wearing a facial covering in compliance with government or corporate mandates regarding health safety protocols. Beyond securing physical locations, face recognition can also be used to manage access to computers, as well as specialized equipment and devices. Overcoming challenges with AI So how did face recognition become so reliable when the technology was once dogged by many challenges, including difficulties with camera angles, certain types of facial expressions, and diverse lighting conditions? Thanks to the emergence of so-called "convolutional" neural network-based algorithms, engineers have been able to overcome these roadblocks. SecurOS FaceX face recognition solution FaceX is powered by neural networks and machine learning which makes it capable of authenticating a wide range of faces One joint effort between New Jersey-based Intelligent Security Systems (ISS) and tech giant Intel has created the SecurOS FaceX face recognition solution. FaceX is powered by neural networks and machine learning which makes it capable of authenticating a wide range of faces and facial expressions, including those captured under changing light, at different resolution levels, and varying distances from the video camera. Secure video management system A common face recognition system deployment begins with IP video cameras that feed footage into a secure video management system connected to a video archive. When the software initially enrolls a person’s face, it creates a "digital descriptor" that is stored as a numeric code that will forever be associated with one identity. The system encrypts and stores these numeric codes in a SQL database. For the sake of convenience and cost savings, the video server CPU performs all neural network processes without requiring any special GPU cards. Unique digital identifiers The next step involves correlating faces captured in a video recording with their unique digital descriptors on file. The system can compare newly captured images against large databases of known individuals or faces captured from video streams. Face recognition technology can provide multi-factor authentication, searching watchlists for specific types of features, such as age, hair color, gender, ethnicity, facial hair, glasses, headwear, and other identifying characteristics including bald spots. Robust encryption SED-compatible drives rely on dedicated chips that encrypt data with AES-128 or AES-256 To support privacy concerns, the entire system features an encrypted and secure login process that prevents unauthorized access to both the database and the archive. An additional layer of encryption is available through the use of Self-Encrypting Drives (SEDs) that hold video recordings and metadata. SED-compatible drives rely on dedicated chips that encrypt data with AES-128 or AES-256 (short for Advanced Encryption Standard). Anti-spoofing safeguards How do face recognition systems handle people who try to trick the system by wearing a costume mask or holding up a picture to hide their faces? FaceX from ISS, for example, includes anti-spoofing capabilities that essentially check for the "liveliness" of a given face. The algorithm can easily flag the flat, two-dimensional nature of a face mask, printed photo, or image on a mobile phone and issue a "spoof" alarm. Increased speed of entry Incorporating facial recognition into existing access control systems is straightforward and cost-effective Incorporating facial recognition into existing access control systems is straightforward and cost-effective. Systems can operate with off-the-shelf security cameras and computers. Users can also leverage existing infrastructure to maintain building aesthetics. A face recognition system can complete the process of detection and recognition in an instant, opening a door or turnstile in less than 500ms. Such efficiency can eliminate hours associated with security personnel checking and managing credentials manually. A vital tool Modern face recognition solutions are infinitely scalable to accommodate global enterprises. As a result, face recognition as a credential is increasingly being implemented for a wide range of applications that transcend traditional access control and physical security to include health safety and workforce management. All these capabilities make face recognition a natural, frictionless solution for managing access control, both in terms of performance and cost.