The notion of cyber-physical power grid is to use metering, communication, and control to make an intelligent and autonomous power system. Therefore, the smart meters form the basis of the cyber-physical system. Recentlyan increasing number of synchrophasor measurement units (PMU) and micro synchrophasor measurement units ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu $ </tex-math></inline-formula> PMU) have been deployed in transmission grids and distribution networks, respectively. Because of the importance of these data, it is imperative to guarantee its integrity and authenticity. In this paper, the measurement structure and data format of the synchrophasor data are analyzed, which reveals the vulnerability of PMU data to cyber-attacks. Then the signal separation method, i.e., independent component analysis algorithm, and its behavior are investigated, which reveals the mechanism of the PMU data attacks. Further, a cognitive radio (CR) based secure network architecture is propose as a countermeasure. With the stochastic but confined choice of the bandwidth and an unfixed number of Splepian tapers, the proposed architecture makes the data stream much harder to intercept by dynamically exploiting the unused PU band. The effectiveness of the proposed network is validated by the simulation results.