Résumé
The use of auctions for the allocation of sponsored search results on the Internet has brought
auctions to the foreground of e-commerce. For instance, for each Google query submitted on the
Internet, there are associated sponsored links displayed to the user, automatically sold by auctions,
resulting in billions of auctions happening online everyday. The sale of sponsored links alone
is responsible for the movement of hundreds of billions of dollars per year. Presently, the vast majority of auctions take place in centralized services, which requires participants to relinquish control to the auctioneer, leaving them at the mercy of the system, with no choice other than to trust the service, hoping that the process is efficient and honest, i.e., that it makes good use of computer resources and won’t disclose private information.
In this thesis, we study whether blockchain technologies, designed for providing trusted environments for the execution of programs (smart contracts), provide an efficient and secure environment for auctions. To make this analysis more concrete, we selected a truthful auction, known as Vickrey-Clarke-Groves for sponsored search (VCG for search), as the basis to assess the impacts of such environments on the auction process.
The first step for our research is the formalization of VCG algorithms, also called mechanisms; we give the specifications of the general VCG mechanism and of its VCG for search variant using the proof assistant Coq, together with the proofs of some of their key properties.
Following up on this sound basis for the definition of auctions, we describe our rationale for the reasoned choice of a sample of blockchain systems to be used as the basis for experimental research. We restricted our systematic and scientifically grounded study of blockchain systems
to Ethereum and Tezos, Ethereum being the industry standard for smart contracts development
and Tezos, a newer system that intends to tackle some of Ethereum’s performance issues by selecting a more efficient consensus mechanism, proof-of-stake (PoS), instead of the proof-of-work approach used in Ethereum. Our performance-comparison methodology focuses on the experimental evaluation of both system in terms of programmability, performance and cost.
Our tests indicate that, at the time of study, Ethereum surpassed Tezos in terms of programmability and community support, which led us to select it as the blockchain of choice to base the rest of our research upon. In addition, our benchmarking provides numerical evidence for the existence of significant scalability limitations for Ethereum and its consensus algorithm; such shortcomings were also noted by the Ethereum community, with scalability solutions being developed in the recent years. Due to their relevancy for the industry, we perform another benchmark comparison between two promising solutions: Polygon PoS and Ethereum’s PoS Merge update. We used our VCG for search smart contract to perform a comparison between standard Ethereum and the scalability updates. This comparison reveals the advantage of both upgrades, but in the case of Polygon, put to light the fact that its dependency on Ethereum’s pricing remains a limitation. Even though performance issues are clearly seen by the community as key to the acceptance of the blockchain technology, the transparency of public blockchains, though it is what ultimately makes the technology secure, presents another severe handicap for the execution of programs that expect a certain level of privacy, such as VCG for search. In the final chapter of this thesis, we analyse the effects of this lack of privacy on the VCG for search ecosystem. We explore some of the industry proposals for privacy, present three new proof-of-concepts variants of VCG for search that increase the privacy of this type of contracts and analyse their effects on the privacy of the auction process as well as their efficiency and monetary impacts on the participants.
Source: http://www.theses.fr/s207280
.
