Real Steps or Not: Auto-Walker Detection in Move-to-Earn Applications

Area of Science:

• Digital Health
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Move-to-Earn (M2E) applications integrate physical activity with digital rewards.
• M2E platforms incentivize real-world movement, unlike Play-to-Earn (P2E) models.
• Increased smartphone use and health consciousness drive M2E adoption.

Purpose of the Study:

• To develop an AI-based method for distinguishing genuine user activity from simulated auto-walker activity in M2E platforms.
• To ensure the integrity of reward distribution mechanisms within M2E applications.
• To validate the model's generalizability across diverse datasets.

Main Methods:

• Utilized six open gait datasets and auto-walker datasets collected via smartphones.
• Developed and evaluated an AI model to discriminate between genuine and simulated gait data.
• Performed unbiased and transparent model evaluation.

Main Results:

• The AI model achieved an F1-score of 0.997 for auto-walker datasets.
• The AI model achieved a perfect F1-score of 1.000 for genuine gait datasets.
• Demonstrated effective discrimination on both seen and unseen datasets, confirming model generalizability.

Conclusions:

• The proposed AI-based method effectively identifies simulated activity in M2E applications.
• This approach enhances the fairness and reliability of M2E reward systems.
• The model's robust performance across diverse datasets supports its practical application in M2E platforms.