Research on a Quantitative Evaluation Model for Privacy Computing Technology Selection

Abstract

Data collaboration in finance and healthcare has become unavoidable, though seldom for lack of competing proposals. In practice, privacy-preserving technology selection remains a half-informed exercise—shaped as much by vendor narratives as by empirical evidence—with the boundaries between Federated Learning (FL) and Secure Multi-Party Computation (MPC) frequently left vague in the literature. Rather than pursuing a theoretically pristine model, this work adopts an engineering stance, grounding its analysis in three pragmatic dimensions: privacy strength, model utility, and system overhead. By drawing on empirical benchmarks from top-tier venues instead of absolute performance predictions, the study comparatively examines FL and MPC across credit scoring and medical image segmentation tasks. The findings resist a universal optimum: MPC delivers robust security at a punitive cost, rendering it economically fragile for deep learning pipelines, whereas FL proves resilient under data heterogeneity yet remains exposed by its semi-honest assumptions. From these trade-offs, a visual decision tree is derived—intentionally coarse, but usable—to anchor technology selection amid conflicting business and compliance constraints.