[Paper Review] Face Morphing Attack Generation & Detection: A Comprehensive Survey

Sushma Venkatesh, Raghavendra Ramachandra+2 more|arXiv (Cornell University)|Nov 3, 2020

Face recognition and analysis94 references17 citations

TL;DR

This comprehensive survey systematically reviews face morphing attack generation and detection techniques, analyzing morphing methods, detection algorithms, benchmarking datasets, and performance metrics. It identifies key challenges such as image quality, aging effects, and user-friendliness, and calls for standardized evaluation frameworks to improve robustness in real-world biometric systems.

ABSTRACT

The vulnerability of Face Recognition System (FRS) to various kind of attacks (both direct and in-direct attacks) and face morphing attacks has received a great interest from the biometric community. The goal of a morphing attack is to subvert the FRS at Automatic Border Control (ABC) gates by presenting the Electronic Machine Readable Travel Document (eMRTD) or e-passport that is obtained based on the morphed face image. Since the application process for the e-passport in the majority countries requires a passport photo to be presented by the applicant, a malicious actor and the accomplice can generate the morphed face image and to obtain the e-passport. An e-passport with a morphed face images can be used by both the malicious actor and the accomplice to cross the border as the morphed face image can be verified against both of them. This can result in a significant threat as a malicious actor can cross the border without revealing the track of his/her criminal background while the details of accomplice are recorded in the log of the access control system. This survey aims to present a systematic overview of the progress made in the area of face morphing in terms of both morph generation and morph detection. In this paper, we describe and illustrate various aspects of face morphing attacks, including different techniques for generating morphed face images but also the state-of-the-art regarding Morph Attack Detection (MAD) algorithms based on a stringent taxonomy and finally the availability of public databases, which allow to benchmark new MAD algorithms in a reproducible manner. The outcomes of competitions/benchmarking, vulnerability assessments and performance evaluation metrics are also provided in a comprehensive manner. Furthermore, we discuss the open challenges and potential future works that need to be addressed in this evolving field of biometrics.

Motivation & Objective

• To provide a systematic overview of face morphing attack generation and detection techniques.
• To establish a stringent taxonomy for morph attack detection (MAD) algorithms.
• To evaluate public databases and benchmarking results for reproducible MAD algorithm testing.
• To identify open challenges in morphing vulnerability, detection robustness, and standardization.
• To guide future research by outlining key issues such as aging, image quality, and user convenience in real-world deployment.

Proposed method

• The paper conducts a comprehensive survey of face morphing techniques, including both holistic and partial face morphing.
• It classifies morph attack detection (MAD) methods into two categories: Still-Morph Attack Detection (S-MAD) and Dynamic-Morph Attack Detection (D-MAD).
• The study evaluates performance using ISO/IEC standard metrics and analyzes results from public competitions and benchmarking efforts.
• It examines the impact of co-variates such as age, gender, ethnicity, image quality, and post-processing on morphing vulnerability and detection.
• The authors analyze existing public databases and assess their suitability for benchmarking MAD algorithms.
• It proposes a structured framework for future evaluation, emphasizing real-world deployment constraints and standardization needs.

Experimental results

Research questions

• RQ1 How do different morphing techniques generate high-quality, indistinguishable morphed face images?
• RQ2 What are the key differences and performance trade-offs between S-MAD and D-MAD detection techniques?
• RQ3 How do image quality, aging, and post-processing affect the vulnerability of face recognition systems to morphing attacks?
• RQ4 What are the current limitations in benchmarking and standardization of MAD algorithms?
• RQ5 How can user-friendly, deployable MAD systems be designed for real-time applications like ABC gates?

Key findings

• The paper identifies that morphing attacks pose a severe threat to eMRTD and passport systems by enabling dual verification with a single morphed image.
• High-quality morphs generated via deep learning techniques are increasingly difficult to detect, especially when partial face regions (e.g., eyes, nose) are targeted.
• D-MAD techniques are more sensitive to imaging conditions such as illumination and occlusions (e.g., glasses, hair), while S-MAD is more affected by facial co-variates like age and ethnicity.
• Despite progress, no standardized vulnerability metric exists for evaluating FRS susceptibility to morphing attacks, highlighting a critical research gap.
• The study reveals that face beautification and digital image submission processes increase morphing risks, especially in countries allowing online passport applications.
• Current benchmarking relies on ISO/IEC metrics, but harmonization and real-world validation across diverse co-variates remain essential for future standardization.