Distributed detection without fusion center (FC) allows each sensor node to exchange measurements with its neighbors and to reach agreement on the existence of the target using the consensus algorithm. Comparing with the centralized system, the decentralized system has better performance, in terms of robustness, scalability and invulnerability. Due to the decentralized and open properties, the security of decentralized system is challenging. In this paper, we introduce the system model of consensus-based distributed detection, and analyze the steady-state detection performance in the presence of the sensing data falsification attack. We define a probabilistic inverse attack model to characterize the malicious behavior. Then, we derive the closed-form expressions for the performance of the consensus-based distributed detection under the defined attack model, in terms of the false-alarm probability, the detection probability and the error probability. Moreover, using the deflection coefficient as the objective function, we derive the blind condition of making the detection system invalid. Simulation results illustrate the effects caused by sensing data falsification attack.