FreqyWM: Frequency WaterMarking for the New Data Economy

May 13, 2024

MLEDGE has already started producing relevant papers in top CS venues! This month our colleague Devriş İşler will be presenting the paper “FreqyWM: Frequency WaterMarking for the New Data Economy“ in IEEE ICDE Conference (ranked A* according to CORE2023) in Utrech, Netherlands. This paper relates to the FedWM research component of the project.

This paper presents a novel technique for modulating the appearance frequency of a few tokens within a dataset for encoding an invisible watermark that can be used to protect ownership rights upon data. It develops optimal as well as fast heuristic algorithms for creating and verifying such watermarks. It also demonstrates the robustness of our technique against various attacks and derives analytical bounds for the false positive probability of erroneously “detecting” a watermark on a dataset that does not carry it. This technique is applicable to both single dimensional and multidimensional datasets, is independent of token type, and can be used in a variety of use cases that involve buying and selling data in contemporary data marketplaces.