New NIST guide explains how to detect morphed images

Face morphing software has the capability to blend two individuals’ photos into a single image, enabling potential identity fraud at secure locations such as buildings, airports, and borders. These morphed images can deceive facial recognition systems, allowing one person to impersonate another. The software is widely accessible, with morphs easily created using mobile applications, desktop graphics programs, or AI tools. While some tools produce high-quality results, others may leave detectable signs, such as inconsistent skin tones or unnatural features around the eyes, nose, lips, or eyebrows.

To combat this issue, the National Institute of Standards and Technology (NIST) has released new guidelines aimed at helping organisations implement detection tools to identify morph attacks before they occur. The publication, titled Face Analysis Technology Evaluation (FATE) MORPH 4B: Considerations for Implementing Morph Detection in Operations (NISTIR 8584), simplifies the concept of morphing and provides actionable advice for organisations, particularly in settings like passport offices and border crossings. The guidelines differentiate between two detection scenarios: single-image morph attack detection, which relies on a questionable photo alone, and differential morph attack detection, which compares a questionable photo with a trusted one. Each method has its advantages and limitations, with recommendations for a combination of automated tools, human review, and clear procedures for handling flagged images.