Ensuring privacy-preserving inference on cryptographically secure data is a well-known computational challenge. To alleviate the bottleneck of costly cryptographic computations in non-linear activations, recent methods have suggested linearizing a targeted portion of these activations in neural networks. This technique results in significantly reduced runtimes with often negligible impacts on accuracy. In this paper, we demonstrate that such computational benefits may lead to increased fairness costs. Specifically, we find that reducing the number of ReLU activations disproportionately decreases the accuracy for minority groups compared to majority groups. To explain these observations, we provide a mathematical interpretation under restricted assumptions about the nature of the decision boundary, while also showing the prevalence of this problem across widely used datasets and architectures. Finally, we show how a simple procedure altering the fine-tuning step for linearized models can serve as an effective mitigation strategy.
Mitigating Disparate Impact of Differentially Private Learning through Bounded Adaptive Clipping
Department of Computer Science, University of Helsinki（计算机科学系，赫尔辛基大学）；Department of Electrical and Computer Engineering, American University of Beirut（电气与计算机工程系，贝鲁特美国大学）
DP-SGD vs PATE: Which Has Less Disparate Impact on Model Accuracy?
Panjab University（旁遮普大学）；OpenMined（OpenMined组织）；Manipal Institute of Technology（马尼帕尔理工学院）；Carnegie Mellon University（卡内基梅隆大学）；IIT Jodhpur（印度理工学院焦特布尔分校）；Ford Motor Company（福特汽车公司）；Innopolis University（因诺波利斯大学）；University of California, San Diego（加利福尼亚大学圣迭戈分校）；University of Oxford（牛津大学）
Federated Inference: Toward Privacy-Preserving Collaborative and Incentivized Model Serving
Department of Electrical Engineering and Computer Science, University of Stavanger（电子工程与计算机科学系，斯塔万格大学）；Department of Data Science, The Catholic University of Korea（数据科学系，韩国天主大学）
