The safe and reliable operation of traction power systems that power railways are crucial to the uninterrupted functioning of this critical public infrastructure. In modern times, traction power systems and railways, in general, are seeing increasing penetration of information and communication technologies (ICT). Traction power systems, like smart grids, have reactive power compensation mechanisms which are controlled remotely through ICT channels. ICT channels are inherently vulnerable to cyber attacks, rendering reactive power compensation mechanisms vulnerable. Malicious reactive power settings through a cyber attack can either hamper the voltage profiles of the traction power system or disrupt the efficient operation, resulting in losses and unsafe operation. In this paper, such attack scenarios are investigated in detail and a methodology is developed to detect such attacks. The proposed methodology is based on detection metrics that are a function of electrical quantities in both train and traction power systems. The effectiveness of these metrics to classify attacks from normal scenarios is justified along with implementation details. The proposed detection method is computationally inexpensive, easy to implement, and reliable when tested using simulations on an Autotransformer Traction Power System model.