Phasor Measurement Unit (PMU) has become a critical component for the modern distribution network, as it records high-resolution synchrophasor data which contain abundant static and dynamic information of the system. However, PMUs are vulnerable to potential cyberattacks, for example, data spoofing attacks. A deliberate PMU spoofing attack can confuse the existing data source authentication models, especially when the models are used for identifying multiple PMUs at the same time. This paper proposes a data-driven cybersecurity defence model which can identify the source information of a large group of PMUs with high accuracy. The model utilizes the inherent correlations among PMUs with a deep neural network to enhance the data source authentication performance. The effectiveness of the proposed model is examined by the PMU data collected from a real distribution network with different error metrics. Through comprehensive numerical experiments, the proposed model provides consistent superior performance in comparison with other state-of-the-art data source identification approaches.