Year: 2025
Type: dissertation
Abstract: Private information retrieval (PIR) schemes allow for the retrieval of information from a database without the database administrator knowing the identity of the database element that was queried. This bachelor’s thesis aims to assess the feasibility of number-theoretic, single-server, computationally private information retrieval (cPIR) by conducting a literature review and comparing a pioneering, quadratic residuosity (QR) -based scheme to a state-of-the-art, ring learning with errors (RLWE) -based one. The analysis of existing literature reveals that, though initially doomed as impractical, cPIR schemes have undergone significant development over the years. Current cPIR protocols achieve good communication complexity and maintain reasonable computational costs. Communication complexities have improved from polynomial to sublinear complexity. Computational costs are still fairly high, but this is to be expected due to the nature of PIR, as databases can be very large. A trivial way to reduce these costs is by confining the privacy requirement to a smaller subset of the database. Moreover, it is generally assumed that a reasonable computational overhead is not a problem in exchange for strong privacy guarantees. The arrival of lattice-based solutions has also improved the privacy of cPIR by offering post-quantum security. When considering the broader adoption of cPIR, factors such as additional safety parameters and situation-specific needs should be considered, beyond just communicational and computational efficiency. Finally, awareness of the existence of cPIR and the question of whether stakeholders are willing to invest in solutions that prevent data collection should be reflected on. In conclusion, single-server cPIR shows strong potential as a practical and feasible solution for database-related privacy.
Source: Aaltodoc (Aalto University)
Author: Korhonen-Chamorro, Ann-Marie
Institution: Aalto University
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7010142643
