Natatsuka, A., Iijima, R., Watanabe, T., Akiyama, M., Sakai, T., & Mori, T. (2022). Understanding the behavior transparency of voice assistant applications using the ChatterBox framework. RAID '22: Proceedings of the 25th International Symposium on Research in Attacks, Intrusions and Defenses, 143-159. https://doi.org/10.1145/3545948.3545970
This paper introduces the ChatterBox framework, developed to analyze the behavior of voice assistant (VA) applications and address privacy concerns stemming from their cloud-based nature. Using natural language processing, ChatterBox generates and analyzes dialogues to understand how VA apps acquire personal information. It supports both English and Japanese and extracts over twice as many dialogues compared to SkillExplorer. Analysis of VA applications revealed that 5–15% collected personal data in non-transparent ways, and 76–94% lacked proper privacy policies. The authors propose improvements for enhancing transparency in VA platforms.
Comentarios