Cyber Physical Systems (CPS) are more vulnerable to cyber and physical attacks than the conventional systems because of the tight integration of the cyber, communication and control aspects of CPS. Deceptive attacks are very severe on such systems because they go undetected when classical detectors like chi-squared detector is used. Once such deceptive attack that we focus in this paper are false data injection attacks, where the attacker inputs carefully designed false attack vectors. Chi-square detector which detects based on the statistics of deviations of the residual i.e. difference of observed measurement and estimated measurement fails to detect these attacks, since the statistics is not changed in these attacks. 'Cosine similarity detector' was proposed to detect False Data Injection (FDI) attacks in smart grid scenario. However, we show that there exists attack vectors that remain stealthy to both chi-square as well as the cosine-square detectors. We propose an random coding scheme to modify the measurements at random, unknown to the attacker while known to the estimator. We establish that our coding scheme in conjunction with the cosine similarity detector provides promising detection performance. © 2019 IEEE.