As one type of advanced sensor, Phasor Measurement Units (PMUs) provide network operators with high-precision time-series Synchrophasor Data (SD) which characterise the operational condition of power systems. Source information of SD acts as a critical role in many PMU-based applications and it is potentially subjected to malicious data spoofing attacks due to the lack of security mechanism in the widely adopted SD specifications. Considering the imperative need for defending against such spoofing attacks, this paper proposes a novel SD Source Authentication (SDSA) scheme by sufficiently exploiting multifractal coupling correlations of SD at multiple locations. Without the need for detailed knowledge of power networks (i.e., system topology and associated parameters) and costly upgrading the existing SD acquisition infrastructure (i.e., PMU hardware and SD transfer/storage device), the proposed method performs model-free and cost-effective SDSA in power systems from a new data-driven perspective of long-range coupling correlation discovery. Specifically, Multiscale Adaptive Coupling Correlation Detrended Analysis (MACCDA) is first developed to reveal the significant coupling multifractal characteristics of time series SD at multiple locations over a broad range of scales simultaneously from which the scale with the most significant multifractality is determined. Then the origin of multifractality and the contribution of SD at each individual location to the overall coupling correlation is quantified by shuffling and surrogating the original SD. Such contribution is further integrated with enhanced Weighted Multifractal Surface Interpolation (WMFSI) to generate synthetic high-resolution SD. Afterwards, distinctive time–frequency signatures are derived from the synthetic SD and they are used by computational intelligence algorithms for SDSA. Simulation results using the real-life SD of Victoria state demonstrate the reliability, efficiency and scalability of the proposed scheme in practical power grids.