提出自适应融合不确定性方法，提升深伪检测在分布外场景下的可靠性。
Architecture-Adaptive Uncertainty Fusion for Deepfake Detection
- 通过优化五类不确定性的加权融合，自动适配不同模型架构。
- 在跨数据集测试中，COF相比随机森林提升7.3倍相关性，且退化更小。
- 无需修改模型，仅需42秒训练，适合实际部署场景。
深伪检测系统在基准测试上接近完美准确率，但真实场景需要可靠的预测不确定性。现有不确定性量化（UQ）方法依赖单一来源，忽略不同模型架构下最优组合方式的差异。本文提出相关性优化融合（COF），一种架构自适应框架，通过在概率单纯形上施加约束优化，最大化融合不确定性分数与预测误差间的皮尔逊相关性，融合了五种互补的不确定性源：认知、随机、校准、容错和分布式。COF无需模型修改，仅需42秒权重优化，远低于5模型深度集成的20–45小时。在FaceForensics++上评估11种架构发现：在相同训练/测试协议下，非线性方法平均相关性（r=0.438）略高于COF（约高5–6%），但在分布偏移时结果逆转。在CelebDF上，COF在11种架构中有9种优于随机森林，最大相关性达0.249（随机森林为0.034）；随机森林跨域性能下降85%至r=0.071，而COF仅下降74%至r=0.116。跨数据集评估（CelebDF与DFDC）显示所有方法均出现灾难性泛化失败：域内相关性0.41–0.47降至近零（平均退化90.7%），其中7种架构出现不确定性反转。结果表明，COF是适用于受控分布部署的实用且可解释框架，并揭示领域自适应UQ是法证应用的核心挑战。
原文摘要 · Abstract (English)
Deepfake detection systems achieve near-perfect accuracy on benchmarks, yet forensic deployment demands reliable prediction uncertainty. Existing uncertainty quantification (UQ) methods rely on single sources and ignore that optimal uncertainty composition varies across architectures. We propose Correlation-Optimized Fusion (COF), an architecture-adaptive framework that fuses five complementary uncertainty sources -- epistemic, aleatoric, calibration, conformal, and distributional -- by maximizing Pearson correlation between fused uncertainty scores and prediction errors via constrained optimization on the probability simplex. COF requires no model modifications and only 42 s of weight optimization, compared to 20--45 h for a 5-model Deep Ensemble. Evaluation across eleven architectures on FaceForensics++ reveals a fundamental trade-off: under matched train/evaluation protocol, non-linear methods achieve approximately 5--6% higher in-domain correlation than COF (mean r = 0.438), but this reverses under distribution shift. On CelebDF, COF outperforms Random Forest in 9/11 architectures with up to 7.3x higher correlation (MaxViT-B: r = 0.249 vs. 0.034); RF degrades 85% cross-domain to r = 0.071, whereas COF retains substantially more signal (74% drop to r = 0.116). Cross-dataset evaluation on CelebDF and DFDC reveals catastrophic generalization failure across all methods: in-domain correlations of 0.41--0.47 collapse to near-zero externally (mean degradation 90.7%), with seven of eleven architectures exhibiting uncertainty inversion. These results establish COF as a practical, interpretable framework for controlled-distribution deployment and identify domain-adaptive UQ as the central open challenge for forensic deployment.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。