Towards Interactive Deepfake Analysis
This work addresses the problem of limited application scenarios in deepfake analysis for researchers and practitioners, though it appears incremental as it builds on existing MLLM techniques.
The paper tackles the limitation of existing deepfake analysis methods by exploring interactive analysis through instruction tuning on multi-modal large language models (MLLMs), resulting in a dataset (DFA-Instruct), benchmark (DFA-Bench), and system (DFA-GPT) as a baseline for the community.
Existing deepfake analysis methods are primarily based on discriminative models, which significantly limit their application scenarios. This paper aims to explore interactive deepfake analysis by performing instruction tuning on multi-modal large language models (MLLMs). This will face challenges such as the lack of datasets and benchmarks, and low training efficiency. To address these issues, we introduce (1) a GPT-assisted data construction process resulting in an instruction-following dataset called DFA-Instruct, (2) a benchmark named DFA-Bench, designed to comprehensively evaluate the capabilities of MLLMs in deepfake detection, deepfake classification, and artifact description, and (3) construct an interactive deepfake analysis system called DFA-GPT, as a strong baseline for the community, with the Low-Rank Adaptation (LoRA) module. The dataset and code will be made available at https://github.com/lxq1000/DFA-Instruct to facilitate further research.