Go to DBLP homepageTCDT: A trust-enabled crowdsourced data trading system in intelligent blockchain over Internet of Things
Abstract: Numerous smart devices connected to Internet of Things (IoT) accelerate rapid development of Crowdsourced Data Trading (CDT) system by capturing quantity of timely data. However, this also arises two key challenges in current CDT system, i.e., the trading security and data quality. Therefore, we propose a trust-enabled CDT system based on blockchain joint with deep reinforcement learning to build a secure and trust trading environment over IoT. Firstly, we propose a Blockchain-enabled two-way Smart Contact (BSC), targeting to maximize benefits of both seller and consumer and make a benefit-cost trade-off under a secure system over IoT. We convert selection process at seller side to ordered knapsack problem and firstly propose the DRL-based Selection (DRLS) scheme through actor-critic algorithm to maximize reward of seller. DRLS is based on a time-limited pointer network and can be adapted to various similar scenarios. Considering data accuracy and coverage, a Quality-based Device Selection (QDS) scheme is designed to find qualified sellers for a consumer. Secondly, we propose a Traceable Trust Computing (TTC) scheme based on historical data backtracking to correct previous trust evaluation, so as to further filter out malicious devices and efficiently improve system security. Meanwhile, an autonomous pricing scheme considering network stability is designed for historical data backtracking, which balances data price and inspires user to report malicious behaviors. Extensive experiments are conducted on Ethereum-based blockchain platform to demonstrate the utility of TCDT. Compared to existing schemes, our TCDT significantly improves data quality by 17.3%, improves security of data trading by 63.9% to 55.7%, and benefits both sellers and consumers in an intelligent IoT.
Loading