Personnel security system of nuclear power plants using machine learning for psychological, behavioral and social media activity analysis.

Abstract

In this research, a novel Personnel Security Model is designed and demonstrated for detecting suspicious people in an organization especially for nuclear power plants. The proposed system composed of three subsystems and a final decision making system such as i. Application software for performing a dynamic questionnaire session of individual employee of the power plant, ii. Eye blink and response time counter for lie detection during the questionnaire session and iii. Another sub system is also introduced for sentiment analysis from social media activities. Then, based on the outputs of three sub-systems, final decision is generated. In first sub-system, According to the science of psychology, suspicious people can be detected by asking some questions, by their response time and their eye blinking, lie can also detected. On the other hand their social media posts can also reflect a person’s actual psychological condition. In this study a person’s answers of the psychological questions, their eye blinking and response time corresponding to the question, and their social media activity are taken in consideration to extract as parameters or features for the final prediction model to find out whether a person is suspicious or not. Experimental results and analysis have been presented to justify the validity of the proposed method.