Enhancing mobile crowd sensing: a blockchain-based decentralized framework with dilated RNN-BiGRU for secure and trustworthy data collection

Abstract

Mobile Crowd Sensing (MCS) systems enable large-scale data collection from heterogeneous IoT and mobile devices but face critical challenges related to data reliability, participant trust, and decentralized validation. Existing blockchain-based MCS frameworks often rely on energy-intensive or static consensus mechanisms and lack adaptive intelligence for detecting malicious contributors, limiting their real-world scalability. This paper proposes an intelligent, decentralized trust management framework that integrates a Delegated Proof-of-Stake (DPoS) blockchain with a Dilated RNN–BiGRU deep learning model. The blockchain ensures tamper-proof transaction validation and trust-based consensus, while the deep network dynamically predicts node reliability using temporal behavior patterns. The integration creates a feedback loop where learned trust scores influence validator selection in real time. The proposed hybrid framework was implemented on a Hyperledger Fabric 2.5 network and evaluated using synthetic MCS data representing heterogeneous environmental, noise, and traffic sensing. The system achieved 98.76% accuracy, 57% latency reduction, and 40% computational cost savings compared with existing PoW- and PoA-based models. These results demonstrate that coupling blockchain consensus with adaptive deep trust modeling can significantly enhance the security, scalability, and efficiency of next-generation MCS systems, making the architecture suitable for real-time, large-scale IoT deployments.