Abstract: Identifying the origins and primary propagation pathways of cascading failures is crucial for preventing the rapid development of cascading failures in the cyber physical power system (CPPS). The coupling between information and physics increases the probability of cross-domain fault propagation and amplifies the risks of fault consequences, but the large-scale transfer of unbalanced power flows remains the inherent driving force behind cascading failures. This paper proposes a high-risk links identification method for cascading failures based on power flow transfer similarity. First, a fault propagation model for the CPPS is constructed to simulate the power flow redistribution and optimization scheduling in the cascading failures. Then, it is revealed that there is a similarity in power flow transfer from faults triggered to cascading fault propagation and system restoration, and direct and indirect power flow transfer similarity calculation methods are proposed to quantify the risk of triggering cascading failures from source faults and the dangerous consequences of power flow transfer from secondary faults, thus achieving comprehensive identification of high-risk segments in cascading failures. Finally, the IEEE 118-bus CPPS and a provincial CPPS are taken as examples for simulation analysis. The simulation results demonstrate that the proposed method can accurately identify high-risk segments in the origin and development process of cascading failures. In the source fault stage, high-risk combinations of faults are more likely to induce cascading failures and lead to major blackouts. In the secondary fault stage, protecting high-risk transmission lines can effectively reduce the scale and losses of cascading failures.