# Micro-Expression-Aware Avatar Fingerprinting via Inter-Frame Feature Differencing

arXiv: https://arxiv.org/abs/2604.23247 · HTML: https://mlsanity.com/paper/2604.23247/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 31 references that are in the map.
- First posted on arXiv 25 Apr 2026.

## Builds on (most cited first, 19 of 19)

- [Supervised Contrastive Learning](https://mlsanity.com/paper/2004.11362/) (2020)
- [One-Shot Free-View Neural Talking-Head Synthesis for Video Conferencing](https://mlsanity.com/paper/2011.15126/) (2020)
- [Lips Don't Lie: A Generalisable and Robust Approach to Face Forgery Detection](https://mlsanity.com/paper/2012.07657/) (2020)
- [EMO: Emote Portrait Alive -- Generating Expressive Portrait Videos with Audio2Video Diffusion Model under Weak Conditions](https://mlsanity.com/paper/2402.17485/) (2024)
- [Thin-Plate Spline Motion Model for Image Animation](https://mlsanity.com/paper/2203.14367/) (2022)
- [LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control](https://mlsanity.com/paper/2407.03168/) (2024)
- [Latent Image Animator: Learning to Animate Images via Latent Space Navigation](https://mlsanity.com/paper/2203.09043/) (2022)
- [ID-Reveal: Identity-aware DeepFake Video Detection](https://mlsanity.com/paper/2012.02512/) (2020)
- [Audio-Visual Person-of-Interest DeepFake Detection](https://mlsanity.com/paper/2204.03083/) (2022)
- [MOL: Joint Estimation of Micro-Expression, Optical Flow, and Landmark via Transformer-Graph-Style Convolution](https://mlsanity.com/paper/2506.14511/) (2025)
- [Avatar Fingerprinting for Authorized Use of Synthetic Talking-Head Videos](https://mlsanity.com/paper/2305.03713/) (2023)
- [Generalizable Deepfake Detection with Phase-Based Motion Analysis](https://mlsanity.com/paper/2211.09363/) (2022)
- [Identity Deepfake Threats to Biometric Authentication Systems: Public and Expert Perspectives](https://mlsanity.com/paper/2506.06825/) (2025)
- [Is It Really You? Exploring Biometric Verification Scenarios in Photorealistic Talking-Head Avatar Videos](https://mlsanity.com/paper/2508.00748/) (2025)
- [Unmasking Puppeteers: Leveraging Biometric Leakage to Expose Impersonation in AI-Based Videoconferencing](https://mlsanity.com/paper/2510.03548/) (2025)
- [Beyond Appearance: Transformer-based Person Identification from Conversational Dynamics](https://mlsanity.com/paper/2510.04753/) (2025)
- [Leveraging Avatar Fingerprinting: A Multi-Generator Photorealistic Talking-Head Public Database and Benchmark](https://mlsanity.com/paper/2603.26934/) (2026)
- [FacialMotionID: Identifying Users of Mixed Reality Headsets using Abstract Facial Motion Representations](https://mlsanity.com/paper/2507.11138/) (2025)
- [Investigating Identity Signals in Conversational Facial Dynamics via Disentangled Expression Features](https://mlsanity.com/paper/2510.11223/) (2025)

## Abstract (arXiv)

Avatar fingerprinting, i.e., verifying who drives a synthetic talking-head video rather than whether it is real, is a critical safeguard for authorized use of face-reenactment technology. Existing methods rely on a fixed, non-differentiable landmark extraction stage that prevents the fingerprinting model from being optimized end-to-end from raw pixels. We propose a preprocessing-free system built on a micro-expression-aware backbone operating on raw video frames, with inter-frame feature differencing as the core design principle: consecutive feature maps are subtracted in the learned deep feature space, so that temporally stable appearance dimensions contribute zero to the output while driver-specific motion dynamics are preserved. A controlled ablation on NVFAIR confirms that temporal motion accounts for the large majority of discriminative performance, and that raw appearance features actively degrade identity separation. Both the choice of backbone and the differencing principle are essential: differencing alone is insufficient when applied to a generic encoder, as appearance-dominated features collapse to near-identical representations across adjacent frames, while the micro-expression-aware F5C backbone retains measurable motion variation that the differencing operation can exploit. Without any external preprocessing, our model achieves an overall AUC of 0.877 on NVFAIR and matches or exceeds the landmark-based baseline on the majority of cross-generator pairs.