The author presents a novel challenge-response-based method to detect deepfake audio calls, highlighting the effectiveness of combining human intuition with machine precision in enhancing detection capabilities.
組み合わせた人間とAIの協力により、ディープフェイクの検出を向上させる。
DeepFake-O-Meter v2.0 is an open-source and user-friendly platform that integrates state-of-the-art methods for detecting AI-generated images, videos, and audio, aiming to provide a convenient service for everyday users and a benchmarking platform for researchers in digital media forensics.
A novel deep learning-based approach that exploits three specialized feature extractors to effectively discriminate between real and AI-generated images, demonstrating robust performance against JPEG compression and improved generalization capabilities.
The proposed method leverages the structure of the latent space of StyleGAN, a state-of-the-art generative adversarial network, to learn a lightweight binary classification model for efficient deepfake detection.
ID-Miner, an identity-anchored and artifact-agnostic deepfake detection method, outperforms baseline detectors under the Rebalanced Deepfake Detection Protocol (RDDP) which reduces the distribution shift between genuine and forged videos.