Abstract: To detect the replay attack in cyber-physical systems (CPS), we propose a measurement noise coding scheme based on a system performance index. Firstly, we briefly describe a replay attack. Secondly, we obtain the maximum allowable measurement noise disturbance within a given range of performance change, add it to sensors to encode measurements, and decode the measurements at the filter end. Finally, we use a fading Kalman filter to estimate the performance of the system in real time. Under normal circumstances of the performance change, the performance index becomes fixed after the system reaches stability. When a replay attack occurs, the decoding information and the encoding process are not synchronized. As a result, the system encounters additional measurement noise that exceeds the maximum allowable range. Consequently, the system performance index goes beyond the given range. With this scheme, no system performance scarification takes place in normal situations. At the same time, the occurrence of a replay attack can be effectively detected by observing the changes in the real-time control performance index of the system.