PROJECT TITLE :

Scalable Uncertainty-Aware Truth Discovery in Big Data Social Sensing Applications for Cyber-Physical Systems - 2017

ABSTRACT:

Social sensing could be a new massive data application paradigm for Cyber-Physical Systems (CPS), where a cluster of individuals volunteer (or are recruited) to report measurements or observations about the physical world at scale. A basic challenge in social sensing applications lies in discovering the correctness of reported observations and reliability of data sources while not prior knowledge on either of them. We tend to refer to this problem as truth discovery. Whereas prior studies have made progress on addressing this challenge, two necessary limitations exist: (i) current solutions failed to fully explore the uncertainty side of human reported knowledge, that ends up in sub-optimal truth discovery results; (ii) current truth discovery solutions are largely designed as sequential algorithms that do not scale well to giant-scale social sensing events. In this paper, we tend to develop a Scalable Uncertainty-Aware Truth Discovery (SUTD) theme to deal with the on top of limitations. The SUTD theme solves a constraint estimation drawback to jointly estimate the correctness of reported data and also the reliability of data sources whereas explicitly considering the uncertainty on the reported data. To deal with the scalability challenge, the SUTD is intended to run a Graphic Processing Unit (GPU) with thousands of cores, which is shown to run 2 to 3 orders of magnitude faster than the sequential truth discovery solutions. In evaluation, we have a tendency to compare our SUTD theme to the state-ofthe- art solutions using 3 real world datasets collected from Twitter: Paris Attack, Oregon Shooting, and Baltimore Riots, all in 2015. The evaluation results show that our new scheme considerably outperforms the baselines in terms of both truth discovery accuracy and execution time.