Document Type
Dissertation
Date of Award
8-31-2020
Degree Name
Doctor of Philosophy in Computing Sciences - (Ph.D.)
Department
Computer Science
First Advisor
Guiling Wang
Second Advisor
Qiang Tang
Third Advisor
Baruch Schieber
Fourth Advisor
Roman Vaculin
Fifth Advisor
Jian Yang
Abstract
Traditional crowdsourcing systems, such as Amazon's Mechanical Turk (MTurk), though once acquiring great economic successes, have to fully rely on third-party platforms to serve between the requesters and the workers for basic utilities. These third-parties have to be fully trusted to assist payments, resolve disputes, protect data privacy, manage user authentications, maintain service online, etc. Nevertheless, tremendous real-world incidents indicate how elusive it is to completely trust these platforms in reality, and the reduction of such over-reliance becomes desirable.
In contrast to the arguably vulnerable centralized approaches, a public blockchain is a distributed and transparent global "consensus computer" that is highly robust. The blockchain is usually managed and replicated by a large-scale peer-to-peer network collectively, thus being much more robust to be fully trusted for correctness and availability. It, therefore, becomes enticing to build novel crowdsourcing applications atop blockchains to reduce the over-trust on third-party platforms.
However, this new fascinating technology also brings about new challenges, which were never that severe in the conventional centralized setting. The most serious issue is that the blockchain is usually maintained in the public Internet environment with a broader attack surface open to anyone. This not only causes serious privacy and security issues, but also allows the adversaries to exploit the attack surface to hamper more basic utilities. Worse still, most existing blockchains support only light on-chain computations, and the "smart contract" executed atop the decentralized "consensus computer" must be simple, which incurs serious feasibility problems. In reality, the privacy/security issue and the feasibility problem even restrain each other and create serious tensions to hinder the broader adoption of blockchain.
The dissertation goes through the non-trivial challenges to realize secure yet still practical decentralization (for urgent crowdsourcing use-cases), and lay down the foundation for this line of research. In sum, it makes the next major contributions.
First, it identifies the needed security requirements in decentralized knowledge crowdsourcing (e.g., data privacy), and initiates the research of private decentralized crowdsourcing. In particular, the confidentiality of solicited data is indispensable to prevent free-riders from "pirating" the others' submissions, thus ensuring the quality of solicited knowledge. To this end, a generic private decentralized crowdsourcing framework is dedicatedly designed, analyzed, and implemented.
Furthermore, this dissertation leverages concretely efficient cryptographic design to reduce the cost of the above generic framework. It focuses on decentralizing the special use-case of Amazon MTurk, and conducts multiple specific-purpose optimizations to remove needless generality to squeeze performance. The implementation atop Ethereum demonstrates a handling cost even lower than MTurk.
In addition, it focuses on decentralized crowdsourcing of computing power for specific machine learning tasks. It lets a requester place deposits in the blockchain to recruit some workers for a designated (randomized) programs. If and only if these workers contribute their resources to compute correctly, they would earn well-deserved payments. For these goals, a simple yet still useful incentive mechanism is developed atop the blockchain to deter rational workers from cheating.
Finally, the research initiates the first systematic study on crowdsourcing blockchains' full nodes to assist superlight clients (e.g., mobile phones and IoT devices) to "read" the blockchain's records. This dissertation presents a novel generic solution through the powerful lens of game-theoretic treatments, which solves the long-standing open problem of designing generic superlight clients for all blockchains.
Recommended Citation
Lu, Yuan, "Crowdsourcing atop blockchains" (2020). Dissertations. 1482.
https://digitalcommons.njit.edu/dissertations/1482
