科研成果详情

Peer-to-Peer (P2P) network can be a promising big data application platform. However, efficient data collection in network still faces a huge challenge in security. In this paper, we utilize the idea of machine learning to select trusted data reporter to collect data. The data collection optimization is translated into how to maximize data coverage and minimize the cost under given budget in malicious network, which has rarely been considered in previous studies. Then a Trust-based Minimum Cost Quality Aware (TMCQA) data collection scheme is proposed to perform data collection optimization. The data collection in TMCQA scheme has the following innovations. (1) A trust evaluation mechanism utilizing the idea of machine learning is established to evaluate the trust of the data reporter. Besides, different from the previous strategy, data reporter is taken as the basic unit of data collection instead of data sample which makes the TMCQA scheme is more practical in the P2P network.(2) An optimized data reporter selection strategy is proposed to select optimized reporters based on the three key evaluation indices to improve data collection performance, which are: (a) The trust value of the data reporter. The more credible the reporter, the higher the quality of the collected data. (b) The data coverage ratio for covering the interesting area of the sensing task. (c) The cost of data collection. The data reporter used to have lower cost will be selected to reduce cost. Finally, we verify the validity of the TMCQA scheme proposed in this paper through various experiments. Comparing to Contribution-based with Trust Value Scheme (CNTVS), Random Data Reporter Selection Scheme (RDRSS), and No Time Decay Scheme (TNTDS), with the same budget, the QoS can be improved by 32.21%, 49.39%, 23.68% respectively. The performance of the TMCQA scheme in a malicious P2P network is significantly better than previous strategies.