Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Emergent Morphing Attack Detection in Open Multi-modal Large Language Models

Marija Ivanovska, Vitomir Štruc

TL;DR

The results indicate that multimodal pretraining can implicitly encode fine-grained facial inconsistencies indicative of morphing artifacts, enabling zero-shot forensic sensitivity and highlights new opportunities to develop state-of-the-art MAD systems through targeted fine-tuning or lightweight adaptation, further improving accuracy and efficiency while preserving interpretability.

Abstract

Face morphing attacks threaten biometric verification, yet most morphing attack detection (MAD) systems require task-specific training and generalize poorly to unseen attack types. Meanwhile, open-source multimodal large language models (MLLMs) have demonstrated strong visual-linguistic reasoning, but their potential in biometric forensics remains underexplored. In this paper, we present the first systematic zero-shot evaluation of open-source MLLMs for single-image MAD, using publicly available weights and a standardized, reproducible protocol. Across diverse morphing techniques, many MLLMs show non-trivial discriminative ability without any fine-tuning or domain adaptation, and LLaVA1.6-Mistral-7B achieves state-of-the-art performance, surpassing highly competitive task-specific MAD baselines by at least 23% in terms of equal error rate (EER). The results indicate that multimodal pretraining can implicitly encode fine-grained facial inconsistencies indicative of morphing artifacts, enabling zero-shot forensic sensitivity. Our findings position open-source MLLMs as reproducible, interpretable, and competitive foundations for biometric security and forensic image analysis. This emergent capability also highlights new opportunities to develop state-of-the-art MAD systems through targeted fine-tuning or lightweight adaptation, further improving accuracy and efficiency while preserving interpretability. To support future research, all code and evaluation protocols will be released upon publication.

Emergent Morphing Attack Detection in Open Multi-modal Large Language Models

TL;DR

The results indicate that multimodal pretraining can implicitly encode fine-grained facial inconsistencies indicative of morphing artifacts, enabling zero-shot forensic sensitivity and highlights new opportunities to develop state-of-the-art MAD systems through targeted fine-tuning or lightweight adaptation, further improving accuracy and efficiency while preserving interpretability.

Abstract

Face morphing attacks threaten biometric verification, yet most morphing attack detection (MAD) systems require task-specific training and generalize poorly to unseen attack types. Meanwhile, open-source multimodal large language models (MLLMs) have demonstrated strong visual-linguistic reasoning, but their potential in biometric forensics remains underexplored. In this paper, we present the first systematic zero-shot evaluation of open-source MLLMs for single-image MAD, using publicly available weights and a standardized, reproducible protocol. Across diverse morphing techniques, many MLLMs show non-trivial discriminative ability without any fine-tuning or domain adaptation, and LLaVA1.6-Mistral-7B achieves state-of-the-art performance, surpassing highly competitive task-specific MAD baselines by at least 23% in terms of equal error rate (EER). The results indicate that multimodal pretraining can implicitly encode fine-grained facial inconsistencies indicative of morphing artifacts, enabling zero-shot forensic sensitivity. Our findings position open-source MLLMs as reproducible, interpretable, and competitive foundations for biometric security and forensic image analysis. This emergent capability also highlights new opportunities to develop state-of-the-art MAD systems through targeted fine-tuning or lightweight adaptation, further improving accuracy and efficiency while preserving interpretability. To support future research, all code and evaluation protocols will be released upon publication.
Paper Structure (7 sections, 4 equations, 4 figures, 2 tables)

This paper contains 7 sections, 4 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Methodology
  4. Experimental Setup
  5. Results
  6. Conclusion
  7. Declaration of generative AI and AI-assisted technologies in the writing process

Figures (4)

  • Figure 1: Method overview. We evaluate MLLMs using standardized prompts and metric-based analysis to assess their inherent ability to detect manipulations.
  • Figure 2: Accuracy–efficiency trade-off of evaluated MLLMs. Medium-sized models achieve the best balance, showing optimal forensic sensitivity.
  • Figure 3: Prompt design impact. Complex prompts degrade performance in small and medium models, while larger models benefit from richer semantic guidance.
  • Figure 4: Interpretability Analysis. LLaVA1.6-Mistral-7B's reasoning and attention highlight key facial cues such as symmetry, texture, and boundary consistency.