Non-Invasive Challenge Response Authentication for Voice Transactions with Smart Home Behavior.

Area of Science:

• Smart Home Technology
• Cybersecurity
• Human-Computer Interaction

Background:

• Voice-controlled smart speakers enhance smart home convenience but face usability challenges with current financial transaction authorization.
• Existing methods often require invasive smartphone notifications, disrupting user experience.

Purpose of the Study:

• To propose and evaluate a non-invasive financial transaction authorization mechanism for smart speakers.
• To improve the usability and reduce the response time of voice-activated financial transactions in smart homes.

Main Methods:

• Leveraging presence and light sensor data for authorization.
• Utilizing Coloured Petri Nets for synthetic data generation and real-user data collection.
• Employing Random Forest and LSTM machine learning models for behavior analysis and prediction.
• Implementing a Challenge-Response protocol with Physical Unclonable Functions (PUFs).

Main Results:

• The proposed mechanism significantly reduces average response time for low-value financial transactions from 7.3s to 3.5s.
• The system demonstrated effective non-invasive authorization, maintaining security levels.
• Machine learning models accurately retrieved smart home behavior information.

Conclusions:

• The developed non-invasive authorization scheme enhances smart speaker financial transaction usability.
• Integrating PUFs and sensor data offers a secure and efficient alternative to traditional methods.
• The approach shows potential for widespread adoption in secure smart home ecosystems.