Machine Learning Algorithms Against Hacking Attack and Detection Success Comparison

Power system protection units have got enormous importance with the growing risk of cyber-attacks. To create a sustainable and well-protected system, power system data must be healthy. For that purpose, many machine learning applications have been developed and used for bad data detection. However, each method has got different detection and application process. Methods have superiority over other methods. Although an algorithm can detect some injections easily, the same algorithm can be fail when injection type changed. So methods have got different success results when the injection types changed. For that reason, different injection types are applied on power system IEEE 14 bus system via created special hacking algorithm. PSCAD and python linkage has been used for simulation and detection parts. 3 different injection types created and applied on the system and five different most popular algorithms (SVM, k- NN, LDA, NB, LR) tested. Each algorithm’s performances are compared and evaluated.