Artistic expression and data protection: Balancing aesthetics with data privacy in IoT

Area of Science:

• Computer Science
• Art and Technology

Background:

• Internet of Things (IoT) art installations offer immersive experiences but raise security and privacy concerns.
• Optimizing differential privacy in these installations presents significant challenges.

Purpose of the Study:

• To explore optimizing differential privacy budgets using deep learning in IoT art.
• To develop a strategy that balances user privacy with the aesthetic integrity of the artwork.

Main Methods:

• A deep learning-based differential privacy budget optimization strategy is proposed.
• Budgets are dynamically allocated using deep learning models and an iterative parameter change law.
• A regularization term is used to limit noise and prevent neural network overfitting.

Main Results:

• The proposed strategy improves the model's generalization capacity.
• Experiments show an accuracy difference of less than 0.5% between noisy and original data models.
• The method effectively protects user privacy while maintaining model availability.

Conclusions:

• The developed strategy successfully balances privacy and availability in IoT art installations.
• It ensures the artwork functions as intended, preserving the artist's message.
• This approach enables secure and aesthetically sound interactive art experiences.