A Stealth False Data Attack on State Estimation with Minimal Network Information

Power system state estimation(SE) is more prone to cyber-attacks being a key operation mechanism to maintain the system observability. To strengthen the cyber security of SE and develop efficient attack detection techniques it is required to examine the existing residual based bad data detection(BDD) security mechanism by developing new, intelligent and stealth false data attack(FDA) models which is near practical. To this aim, the presented work focuses on model free deep convolutional generative adversarial networks(DCGAN) based single bus target stealth FDA injection on SE with minimal network information. Against the previous model based methods in this the inherent measurement data distribution and physical model is learnt only by means of the easily available measurement data on the single bus target area. The single bus target area method is the near practical approach and will be difficult to recognize as against the previous tie-line based and whole area attack region. The generated attack is validated on IEEE 14 bus system and the experimental results shows that it exhibits sufficiently high intensity and residual within the predefined threshold as for the normal data and can easily bypass the BDD mechanism with high success rate.