Fraunhofer IGD Tackles Face-Morphing Fraud

Innovative Solutions for Detecting Fake Faces

Fraunhofer IGD is spearheading the development of advanced and reliable methods to detect and prevent face-morphing attacks, which are increasingly posing significant threats to biometric security systems. Face morphing involves merging the features of multiple faces into a single composite image, enabling criminals to create false identities that allow them to cross borders undetected. To address this growing concern, Fraunhofer IGD will present its latest research on face-morphing detection at the it-sa Expo&Congress in Nuremberg, taking place from October 22 to 24.

While face morphing may sound like an advanced version of photoshopping, it presents a serious threat to international security. This technique uses sophisticated algorithms to combine facial features, structures, and patterns from two or more individuals into a seamless image. What may seem like a harmless or even humorous technique has, unfortunately, become a powerful tool for criminals looking to bypass biometric security checks.

Exploiting Vulnerabilities in Biometric Systems

Criminals use this method to alter passport photos, blending their face with that of another individual. The resulting image can then be used to obtain fraudulent identification documents, such as passports or ID cards, which enable them to pass through border controls undetected. “Neither human vision nor current biometric systems can reliably detect this type of sophisticated fraud,” explains Florian Kirchbuchner, a biometrics expert at the Fraunhofer Institute for Computer Graphics Research IGD.

At major travel hubs like airports, most identity checks are performed by automated biometric systems. Criminals have taken advantage of these systems, exploiting weaknesses to slip through the cracks undetected. Morphed images frequently go undetected because biometric systems are designed to accommodate minor variations in an individual’s facial features, Kirchbuchner notes.

Advanced Detection in the ATHENE Project

Fraunhofer IGD is addressing this critical issue by developing more sophisticated detection algorithms capable of staying ahead of the criminals. “Our aim is to develop a thorough face-morphing detection system that allows us to detect fraudulent identities before they can slip through unnoticed,” Kirchbuchner states. As part of the ATHENE project, which is the German National Research Center for Applied Cybersecurity, the research team is using deep learning and artificial intelligence (AI) to study and anticipate potential morphing techniques.

By using generative adversarial networks (GANs), researchers at Fraunhofer IGD can simulate new faces that combine features from two original faces, helping to anticipate possible attack methods and prepare their detection systems. “This allows us to fine-tune our detection technology and develop more robust defences against face morphing,” Kirchbuchner explains.

Harnessing Machine Learning for Security

Within the scope of the ATHENE project, the research team is also focusing on facial image quality control and identity management, ensuring that biometric security systems are capable of withstanding sophisticated attacks. Additionally, Fraunhofer IGD is investigating how biometric systems can be integrated into smaller, embedded systems with limited computing power, such as access control points or even mobile phones.

One exciting application of this research is the use of biometrics in head-mounted displays, such as those found in virtual reality (VR) and augmented reality (AR) systems. These systems could assist border control officers in identifying users, as well as ensuring high levels of security, even with varying eye positions or movements. “By using eye tracking in these headsets, we can accurately identify the wearer, enhancing security and reducing the likelihood of spoofing,” says Kirchbuchner.

Live Demonstration at it-sa Expo&Congress

Fraunhofer IGD will showcase its cutting-edge research at the it-sa Expo&Congress from October 22 to 24 in Nuremberg, Germany. Visitors to Hall 6, Booth 6-314, will have the opportunity to engage with a live demonstrator simulating an airport security scenario. Participants can act as border agents to test their ability to identify morphed passport photos, and even create a morphed photo of themselves using biometric systems.

Florian Kirchbuchner will also be presenting a talk titled “Spoofing Attacks on Biometric Systems. Where Does the Research Stand?”, providing in-depth insights into the challenges presented by face morphing and the solutions being developed to address them.

Biometric Week in Darmstadt