CAREER: Tracking, Revealing and Detecting Crowdsourced Manipulation

点击翻译按钮获取中文摘要

The goal of this project is to create the algorithms, frameworks, and systems for defending the open web ecosystem from emerging threats. This project aims to (i) analyze malicious tasks and behaviors of crowdturfers; (ii) detect malicious tasks on crowdsourcing platforms by developing novel malicious task detectors; (iii) design and build a task blacklist; (iv) uncover the ecosystem of crowdturfers and detect crowdturfers; (v) combine crowdturfer detection approaches with other malicious participants detection approaches. Crowdsourcing systems have successfully leveraged the attention of millions of "crowdsourced" workers to tackle vexing problems. From specialized systems for crisis mapping, for protein folding, for translation to general-purpose crowdsourcing platforms. However, these positive opportunities have sinister counterparts: large-scale "crowdturfing", wherein masses of cheaply paid workers can be organized to spread malicious URLs in social media, formation of artificial grassroots campaigns ("astroturf"), and manipulation of search engines. As a result, crowdsourced manipulation threatens the foundations of the open web ecosystem, reducing the quality of online social media, degrading our trust in search engines, manipulating political opinion and ultimately, reducing security and trustworthiness of cyberspace. Products of the research will be available for public use. The education and outreach efforts of the project are tightly linked to the research goals through curriculum development, workshops, direct training of underrepresented women, and involvement of industry. The intellectual merit of the project is it will advance the current security systems against crowdsourced manipulation in cyberspace. This project fundamentally alters the landscape of malicious task problems by detecting malicious tasks in crowdsourcing platforms. Early detection of malicious tasks has the potential to transform our solutions for secure and trustworthy information systems. Given our malicious task detection systems, identified malicious tasks can be used as samples for creating new blacklists. The blacklists have the potential to prevent propagation of malicious tasks to popular online target sites. Given our novel techniques, detecting almost all crowdturfers becomes a distinct possibility in the near future. Overall, this project will advance knowledge and understanding the crowdsourced manipulation problem, and the proposed detection framework will complement the current security systems against crowdsourced manipulation. The broader impacts of the proposed work include advances to discovery and understanding while promoting teaching, training and learning. To benefit society, the proposed malicious task and crowdturfer detection framework including task blacklists will enable crowdsourcing service providers and target sites providers to detect crowdsourced manipulation with protecting information quality and trust.