While Federated Learning (FL) mitigates direct data exposure, the resulting trained models remain susceptible to membership inference attacks (MIAs). This paper presents an empirical evaluation of Differential Privacy (DP) as a defense mechanism against MIAs in FL, leveraging the environment of the 2025 NIST Genomics Privacy-Preserving Federated Learning (PPFL) Red Teaming Event. To improve inference accuracy, we propose a stacking attack strategy that ensembles seven black-box estimators to train a meta-classifier on prediction probabilities and cross-entropy losses. We evaluate this methodology against target models under three privacy configurations: an unprotected convolutional neural network (CNN, $ε=\infty$), a low-privacy DP model ($ε=200$), and a high-privacy DP model ($ε=10$). The attack outperforms all baselines in the No DP and Low Privacy settings and, critically, maintains measurable membership leakage at $ε=200$ where a single-signal LiRA baseline collapses. Evaluated on an independent third-party benchmark, these results provide an empirical characterisation of how stacking-based inference degrades across calibrated DP tiers in FL.

翻译：尽管联邦学习（FL）可减轻直接数据暴露风险，但训练所得模型仍易受成员推理攻击（MIAs）。本文基于2025年NIST基因组隐私保护联邦学习（PPFL）红队测试环境，对差分隐私（DP）作为MIAs防御机制进行实证评估。为提高推理精度，我们提出一种堆叠攻击策略，该策略集成七个黑盒估计器，通过预测概率与交叉熵损失训练元分类器。我们针对三种隐私配置下的目标模型评估该方法：无保护卷积神经网络（CNN，$ε=\infty$）、低隐私DP模型（$ε=200$）及高隐私DP模型（$ε=10$）。该攻击在无DP与低隐私设置中全面超越所有基线方法，且关键性突破在于：当单信号LiRA基线失效时，该攻击在$ε=200$条件下仍能保持可测量的成员泄露。基于独立第三方基准的评估结果，实证刻画了堆叠推理在FL校准DP层级间的性能衰减特征。