Many location-based services rely on a point-in-polygon test (PiP), checking whether a point or a trajectory lies inside a geographic zone. Since geometric operations are expensive in zero-knowledge proofs, privately performing the PiP test is challenging. In this paper, we answer the research questions of how different ways of encoding zones affect accuracy and proof cost by exploiting gridbased lookup tables under a fixed STARK execution model. Beyond a Boolean grid-based baseline that marks cells as in- or outside, we explore a distance-aware encoding approach that stores how far each cell is from a zone boundary and uses interpolation to reason within a cell. Our experiments on real-world data demonstrate that the proposed distance-aware approach achieves higher accuracy on coarse grids (max. 60%p accuracy gain) with only a moderate verification overhead (approximately 1.4x), making zone encoding the key lever for efficient zero-knowledge spatial checks.
翻译:许多基于位置的服务依赖于点面测试(PiP),即检查一个点或轨迹是否位于地理区域内。由于在零知识证明中几何运算成本高昂,隐私化执行点面测试具有挑战性。本文通过固定STARK执行模型下基于网格的查找表,研究了不同区域编码方式如何影响精度与证明成本这一研究问题。除了将网格单元标记为区域内/外的布尔型基线方法外,我们探索了一种距离感知编码方法:该方法存储每个单元到区域边界的距离,并利用插值在单元内部进行推理。在真实数据上的实验表明,所提出的距离感知方法在粗粒度网格上实现了更高的精度(最高提升60%p),同时仅产生适度的验证开销(约1.4倍),这使区域编码成为实现高效零知识空间检查的关键杠杆。