Year: 2025
Type: dissertation
Abstract: The volume of data that needs to be stored and transferred is increasing. Centralized solutions, i. e., cloud providers offer scalable storage solutions, making data availability and control dependent on a single logical entity. A self-scalable, distributed, and content-centric alternative for data exchange are peer-to-peer (P2P) data networks. P2P data networks can allow users to regain control over their data. The content-centric approach enables users to directly request the content from the network via mostly pseudonymous identifier. Unfortunately, a query reveals interest in specific content and the source of a query can be linked to a user. The combination of source and content of a query reveals the interest of a user, enabling profiling. Furthermore, performance improvements and fail-safe mechanisms spread interests widely through the network. In this thesis, we design and explore different protocols using privacy-enhancing mechanisms. The goal of the privacy enhancement is to improve content discovery privacy with an adjustable overhead in P2P data networks. Specifically, we focus on source and query privacy mechanisms, providing different privacy guarantees. Source privacy aims to obfuscate the source of a query. With network-level methods, we increase the number of possible sources of a query, obfuscating the source, improving source privacy. The modifications to the content discovery process aim to provide plausible deniability up to relationship anonymity for messages. The second focus, query privacy, aims to hide the item of a request. By adding probabilistic data structures and cryptographic protocols to the content discovery process, we obfuscate the item of interest in a query, increasing query privacy. Specifically, we increase the anonymity set of a query from one item to a range of items, increasing the anonymity set ideally to all existing items in the network. Private set intersection reduces the privacy impediments of the content providers. We apply, explain, and evaluate our protocols in the context of a prime example of the current P2P data networks: the InterPlanetary File System (IPFS). The protocols are evaluated using code experiments with a simulated network, analyzing the overhead’s impact on performance. We analyze the protocols in the context of the public IPFS network by using passive measurements to evaluate the protocols’ real-world feasibility. Each proposed protocol improves the privacy of content discovery, by reducing the overall information leak, effectively limiting large-scale attacks. A combination of the underlying mechanisms with additional mechanisms like cover traffic can provide strong protection.
Author: Daniel, Erik
Institution: Technische Universität Dresden
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7112895818
