Differential privacy and artificial intelligence: potentials, challenges, and future avenues
Abstract Privacy preservation has become an increasingly critical concern in applications where data serves as a cornerstone for decision-making and innovation. Researchers and developers are dedicated to identifying and mitigating emerging risks while improving the privacy of existing systems. Arti...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-05-01
|
| Series: | EURASIP Journal on Information Security |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s13635-025-00203-9 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Privacy preservation has become an increasingly critical concern in applications where data serves as a cornerstone for decision-making and innovation. Researchers and developers are dedicated to identifying and mitigating emerging risks while improving the privacy of existing systems. Artificial intelligence technologies can dynamically detect and address privacy concerns. Differential privacy, with its strong and verifiable assurances, is critical for addressing rising concerns about data privacy in the age of big data and advanced analytics. Combining differential privacy with AI has been identified as a solution for balancing data usage for insights while maintaining individual privacy. However, research in this field is still scarce due to the recent widespread application of artificial intelligence in many industries. This paper reviews current literature, professional websites, and other online resources to determine the potential, challenges, and future directions of combining differential privacy with AI. The key opportunities identified in this study include enhancing privacy (reported in 27% of the reviewed papers), promoting responsible AI (21%), facilitating data sharing (14.5%), and minimizing AI model biases (12.5%). Several concerns, however, require additional exploration, including accuracy trade-offs, computational complexity, regulatory restrictions, expertise, data usability, scalability constraints, and bias concerns. Given that this combination is still a relatively new field, AI developers and users need to stay current on differential privacy research and implement appropriate measures. |
|---|---|
| ISSN: | 2510-523X |