[4Xin2] Poster session 2

[4Xin2-96] Proposal for Special Fraud Training Tools Utilizing Generated AI 2

Construction of the Fraud Risk Estimation Model Using Physiological Information and Personality Traits to Realize Individual Feedback

〇Hina Ikeda¹, Kenta Ide¹, Megumi Chikano¹, Kohei Yoshino¹, Takeshi Konno¹, Masayuki Kiriu² (1.Fujitsu Limited., 2.Toyo University)

Keywords:fraud crime, generative AI, personality traits, big five

In Japan, the damage of special fraud is increasing. To reduce fraud damage, we are developing the special fraud training AI tool with a function to experience the conversation of special fraud and a function to feedback the potential fraud damage risk. Therefore, this study aims to construct an AI model to estimate the risk, which can respond to individual feedback. In addition to using physiological information during training and age as the feature quantity, we also constructed a model using personality traits, considering that gullibility is related to personality. As a result of the evaluation using the data of 29 elderly persons, the model construction with higher accuracy was possible by adding the personality traits to the feature quantity, and the individual feedback became possible. In the future, the verification will be carried out by the training in the municipality, and the effectiveness will be confirmed.