Locational Marginal Prices (LMPs) are central to electricity-market operations but are vulnerable to false data injection attacks (FDIAs) that can mislead dispatch decisions and distort prices. Although recent cybersecurity research has mainly focused on modelling profit-motivated attacks and analysing their financial impact, far less attention has been paid to developing robust, data-driven defence models. A key barrier is the lack of realistic, open datasets that capture stealthy attack behavior grounded in power system physics. To address this gap, we develop the Stealthy Manipulated LMP Timeseries (SMLT) dataset framework; an open-source benchmark for studying FDIAs on electricity markets. SMLT models eight different attack cases including attacks on topology, demand profiles, system parameters, and transmission limits. SMLT provides 20 weeks of hourly LMP timeseries with ground-truth labels, enabling detailed spatio-temporal analysis and facilitating the development of detection algorithms. We present statistical insights into manipulated LMP behaviors and demonstrate the dataset’s utility through a case study using a statistical anomaly detection method. SMLT aims to support reproducible, physics-aware research on securing electricity markets against evolving cyber threats.