Ditto : Towards Decentralised Similarity Search for Web3 Services

Keizer, Navin and Ascigil, Onur and Król, Michał and Pavlou, George (2023) Ditto : Towards Decentralised Similarity Search for Web3 Services. In: 2023 IEEE International Conference on Decentralized Applications and Infrastructures (DAPPS) :. 2023 IEEE International Conference on Decentralized Applications and Infrastructures (DAPPS) . IEEE, pp. 66-75. ISBN 9798350335361

Text (Ditto: Towards Decentralised Similarity Search for Web3 Services) - Accepted Version
Available under License Creative Commons Attribution.
Download (0B)

Text (Ditto: Towards Decentralised Similarity Search for Web3 Services) - Accepted Version
Available under License Creative Commons Attribution.
Download (0B)

Text (Ditto: Towards Decentralised Similarity Search for Web3 Services) - Accepted Version
Available under License Creative Commons Attribution.
Download (0B)

Text (Ditto: Towards Decentralised Similarity Search for Web3 Services) - Accepted Version
Available under License Creative Commons Attribution.
Download (0B)

Text (Ditto: Towards Decentralised Similarity Search for Web3 Services)
main.pdf - Accepted Version
Available under License Creative Commons Attribution.
Download (796kB)

Abstract

The Web has become an integral part of life, and over the past decade, it has become increasingly centralised, leading to a number of challenges such as censorship and control, particularly in search engines. Recently, the paradigm of the decentralised Web (DWeb), or Web3, has emerged, which aims to provide decentralised alternatives to current systems with decentralised control, transparency, and openness. In this paper we introduce Ditto, a decentralised search mechanism for DWeb content, based on similarity search. Ditto uses locality sensitive hashing (LSH) to extract similarity signatures and records from content, which are stored on a decentralised index on top of a distributed hash table (DHT). Ditto uniquely supports numerous underlying content networks and types, and supports various use-cases, including keyword-search. Our evaluation shows that our system is feasible and that our search quality, delay, and overhead are comparable to those currently accepted by users of DWeb and search systems.