| Title |
Local Differential Privacy-based Chat Data Privacy Protection Technology |
| DOI |
https://doi.org/10.5573/ieie.2024.61.1.34 |
| Keywords |
Personal information protect; Differential privacy; Local differential privacy; Gaussian mechanism |
| Abstract |
In recent times, as businesses have expanded their utilization of vast amounts of data to extract valuable information, concerns have arisen regarding security measures such as data encryption and anonymization for personal information. In particular, integrating new chat data from users into models through AI chatbots can lead to differences between the input and output, potentially resulting in data leaks and vulnerabilities that could be exploited to identify individuals. Moreover, even when users trust the server, they may still have concerns about how their information included in chat data is disclosed and used. While data encryption can be employed for privacy protection, the risk of service administrators having unauthorized access remains. Therefore, this paper proposes a method of applying Gaussian Local Differential Privacy (GLDP) to customer chat data collected through AI chatbots. By integrating GLDP into existing chat services, it is possible to preserve user data privacy while providing an advantageous environment for data analysis. In the latter part of this paper, the proposed algorithm's completeness is verified through model implementation and experiments, and comparisons are made regarding ε, privacy levels, data accuracy, and more. |