Developments in Deepfake Detection and Face Recognition

The field of deepfake detection and face recognition is rapidly advancing, with a focus on improving the robustness and generalizability of detection models. Recent research has explored the use of novel approaches, such as leveraging intermediate features of vision transformers, multimodal datasets, and reconstruction-based methods, to detect and prevent deepfake attacks. Additionally, there is a growing interest in developing more effective out-of-distribution detection strategies and improving the adversarial robustness of AI-generated image detectors. Noteworthy papers include Logits-Based Finetuning, which proposes a reconstruction-based method for out-of-distribution detection, and AuthGuard, which incorporates language guidance to improve deepfake detection generalization. Overall, the field is moving towards more robust and adaptable solutions to counter the evolving threat of deepfakes and improve face recognition systems.

Sources

Leveraging Intermediate Features of Vision Transformer for Face Anti-Spoofing

SEAR: A Multimodal Dataset for Analyzing AR-LLM-Driven Social Engineering Behaviors

Logits-Based Finetuning

TalkingHeadBench: A Multi-Modal Benchmark & Analysis of Talking-Head DeepFake Detection

Enhancing Abnormality Identification: Robust Out-of-Distribution Strategies for Deepfake Detection

NTIRE 2025 XGC Quality Assessment Challenge: Methods and Results

DFBench: Benchmarking Deepfake Image Detection Capability of Large Multimodal Models

RAID: A Dataset for Testing the Adversarial Robustness of AI-Generated Image Detectors

Improving Out-of-Distribution Detection with Markov Logic Networks

Towards Large-Scale Pose-Invariant Face Recognition Using Face Defrontalization

AuthGuard: Generalizable Deepfake Detection via Language Guidance

Practical Manipulation Model for Robust Deepfake Detection

Built with on top of