P²ESA: Privacy-Preserving Environmental Sensor-Based Authentication

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Internet of Things (IoT) devices are increasingly prevalent in daily life.
• IoT data can be used for user identification and authentication.
• Collected data may inadvertently reveal sensitive personal information beyond authentication needs.

Purpose of the Study:

• To develop a method for removing privacy-revealing information from IoT data.
• To maintain high user authentication performance after privacy enhancement.
• To address computational limitations of edge devices in machine learning pipelines.

Main Methods:

• Utilized adversarial deep learning techniques for privacy information obfuscation.
• Developed offloading techniques for computationally weak edge devices.
• Implemented and trained the P²ESA system on multimodal IoT datasets.

Main Results:

• Achieved user identification rates between 75.85% and 93.31%, significantly outperforming the 6.67% baseline.
• Demonstrated efficient deployment and training of the P²ESA system.
• Successfully obfuscated sensitive user information while preserving authentication utility.

Conclusions:

• P²ESA offers a promising solution for secure authentication in IoT environments.
• The method effectively balances user identification accuracy with robust privacy protection.
• P²ESA addresses computational challenges for edge devices in machine learning applications.