A unique data-driven methodology for system state estimation against attacks from bogus data injection that cannot be observed is provided in this study. The proposed system continuously detects and classifies attacks involving false data injection. Using the knowledge that has been learned, the control signal is then retrieved. This method is carried out by three main modules that implement state-of-the-art detection categorization and retrieval of control signals. The detecting module keeps track of any deviation pattern brought on by a complicated plane strike by tracking historical phasor measurement changes. The direction, amount, and ratio of the fake data injection are among the assault elements revealed by this method. With the use of this knowledge, the signal recovery module can swiftly obtain the actual control signal and get rid of the false data that was purposefully added. The classifier module can be used to learn more details about the attack type.