A Reinforcement Learning-based Task Classification Mechanism for Privacy-Enhanced Mobile Crowdsensing Strategy
Mengyao Peng, Hui Lin, Xiaoding Wang · 2021
The emergence of the Internet of Things enables efficient connections between things through the Internet, providing a professional platform for information collection, transmission, and sharing. Nowadays, as an important computing model in the Internet of Things, Mobile Crowdsensing(MCS) has received more and more attention. It provides strong technical support for the collection and interaction of information between individuals or devices from different regions. However, while realizing data sharing, it also inevitably brings about the privacy leakage of related data. In order to solve this problem, many privacy protection strategies based on different technologies have been proposed to ensure the privacy of crowdsensing tasks and crowdsensing data. They include the strategies for classifying and grading crowdsensing tasks and workers. In response to this strategy, this paper proposes a algorithm to calculate the number of classifications of the crowdsensing tasks and workers to improve classification efficiency, that is, a reinforcement learning-based task classification mechanism(RTCM). This mechanism uses the Q learning algorithm in reinforcement learning. Through continuous learning iterations, it is possible to select strategies with higher privacy protection degree and task completion quality from different classification strategies. In this way, the system can implement a more efficient privacy protection function according to the optimal strategy. Experiments show that this mechanism can improve the efficiency of crowdsensing tasks and workers classification. And, in different areas, according to different demand standards, the appropriate classification strategy can be quickly selected.