Keizer, Navin V.; (2023) Decentralising Content Retrieval on the Decentralised Web. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Keizer__thesis.pdf - Other Download (4MB) | Preview

Abstract

In recent times, the control, governance, and management of the Web have become increasingly centralised, which has led to several challenges such as a lack of security and privacy protection, as well as increased censorship. To overcome these issues, a number of initiatives have emerged that offer decentralised counterparts of various components of the Web, aiming to create a fully decentralised Web, also known as DWeb or Web3. Novel technologies like blockchains and decentralised storage networks (DSN) offer ways of establishing trust and storing content without centralised trust assumptions, while established technologies like peer-to-peer (P2P) overlay networks are used as their base. This thesis explores the feasibility of decentralising the Web from a content retrieval perspective. In order to understand the emerging paradigm of the DWeb, a framework is proposed for studying novel works in the area, and an extensive analysis is provided on current initiatives. The area of content search has been most unexplored, and therefore this thesis presents a truly decentralised search engine based on similarity search, which can be extended to implement keyword search. This mechanism achieves up to 57% recall of results compared to baseline, and achieves sub millisecond delays in keyword search for a network size up to 5000. Furthermore, in order to facilitate a DWeb, resource sharing over P2P networks requires fair exchange of work for resources. To realise this, a decentralised allocation mechanism is proposed on the blockchain based on the stable matching problem. This system achieves low smart contract costs, reducing cost by at least 52% compared to prior work. Furthermore, allocation throughput and delays improve over know auction mechanisms, staying under 0.5 seconds for 9000 clients. To further facilitate this system, trust and reputation systems are required, and therefore a personalised reputation system framework is proposed, using a model learning from previous blockchain transactions and user preferences. Finally, novel DWeb technologies generally leverage P2P networking as their base, and a number of known challenges remain in this area such as security and privacy concerns. One main challenge is presented by networking infrastructures such as network address translators (NAT), which limit individuals ability to participate in P2P networks. To tackle this issue, a decentralised NAT traversal mechanism is proposed using blockchain smart contracts for resource sharing, and reputation-based peer discovery. Evaluation of the system shows that using a reputation system of combined metrics and a two round peer discovery is able to achieve well under 5% malicious nodes chosen as NAT providers, even with the number of malicious nodes in the network reaching 30%, and the system is able to stabilise in 5-10 service cycles.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item