Detection of False Data Injection Attacks in Power Systems Using Deep Autoencoder with Attention Mechanism

The evolution of electrical power systems into smart grids has increased their efficiency but has also introduced new vulnerabilities, particularly in the form of cyber-attacks. One such threat is the False Data Injection (FDI) attack, which can bypass traditional Bad Data Detection (BDD) mechanisms, thereby posing significant risks to system security. Detecting such attacks is crucial to ensure the security and stability of power systems. In this paper, a self-supervised deep autoencoder model with an attention mechanism is applied to detect FDI attacks in power systems. Two datasets were used: one containing real measurements from an IEEE 14-bus system and another containing attack vectors. These datasets were randomly combined to form a compromised dataset. A model was then developed with the goal of distinguishing real measurements from attack vectors. The performance of the model was subsequently analyzed using appropriate metrics, which achieved a high detection rate with minimal false positives, demonstrating its robustness