In many sensor network applications, sensor nodes are deployed in unattended environments, and hence are vulnerable to physical attacks, potentially compromising the node’s cryptographic keys. False sensing reports can be injected through compromised nodes, which can lead to not only false alarms but also the depletion of limited energy resources in battery powered networks. Ye et al. [4] proposed the statistical en-route filtering scheme to detect and drop such false reports during the forwarding process. Since each node in this scheme has a limited amount of information for verification, the detection power is largely affected by the choice of routing paths. In this paper, a path selection method is proposed, to improve the detection power of the statistical filtering. Each node evaluates the detection power of each incoming path from the base station and chooses the most secure path for data delivery against false data injection attacks. The effectiveness of the proposed method against false data injection attacks is shown via simulation results.