CNN-RL: A Deep Learning–Based Adaptive Crowdsourcing Model for Cybersecurity Education

Abstract

Cybersecurity education needs practical and up-to-date training exercises yet creating and updating them quickly is difficult as new threats appear. This paper presents CNN-RL, an adaptive crowdsourcing framework that uses convolutional neural networks (CNN) to review experiment quality and reinforcement learning (RL) to adjust rewards and tasks automatically. The system links three groups such as Constructors who design experiments, Demanders who define learning needs and users who test the materials. It works through two channels that share feedback and distribute rewards using the Forward Incentive Model (FIM) and the Backward Incentive Model (BIM). A case study at Guangzhou University shows that CNN-RL increased student participation and doubled the number of experiments created. The model helps educators and professionals build cybersecurity exercises that stay current, fair and engaging. Future work will focus on improving model rules and testing CNN-RL in more institutions.