Resilient Network Embeddings for Friend-to-Friend Networks
Final Report Abstract
The internet plays a major role in today’s communication. However, access to the internet can be abused by malicious actors to surveil citizens or to block access to important information. Resisting such practices requires anonymous communication networks, which not only hide the content of communications, but also hide metadata and resist censorship. One direction that such networks can take is that of friend-to-friend networks: In such networks, only peers that mutually trust each other form a ‘link’, such that messages only travel over a series of trusted links. Building friend-to-friend networks in practice however is challenging, as constructing routing algorithms that work on a decentralized, peer-to-peer network in a privacy-preserving manner is hard. In our research, we focused on embeddings as a good candidate to build routing algorithms to be used in friend-to-friend networks. At first, we used an embedding-based routing approach to construct a fast peer-to-peer network. The team extended this to implement payment networks, which are a specific group of friend-to-friend networks. Then, we took a closer look at the different kinds of privacy, and how they can be formalized. In doing so, a clear hierarchy of privacy notions has been developed. Continuing our friend-to-friend network research, we investigated how tree-based embeddings can be implemented in a secure manner, and hence studied the problem of leader election in peer-to-peer networks. There, we have shown that the simple ‘three-majority-voting’ method provides a good candidate for friend-tofriend networks. After an interruption of the project caused by the Covid pandemic and the relocation of the research group, the focus has shifted away from friend-to-friend network to mix networks as a method of anonymization. This was motivated by results that show that the anonymity in peer-to-peer networks is likely less than usually assumed, as the anonymity sets get rather small. In this area, we focused on the improvement of group messaging, as existing techniques put the work of sending a message to multiple group members on the clients. We developed a technique which allows the client to instead offload replication of messages to a server, while retaining the anonymity guarantees of a mix network.
Publications
-
On Privacy Notions in Anonymous Communication. Proceedings on Privacy Enhancing Technologies, 2019(2), 105-125.
Kuhn, Christiane; Beck, Martin; Schiffner, Stefan; Jorswieck, Eduard & Strufe, Thorsten
-
Attack resistant Leader Election in Social Overlay Networks by Leveraging Local Voting. Proceedings of the 21st International Conference on Distributed Computing and Networking, 1-10.
Byrenheid, Martin; Strufe, Thorsten & Roos, Stefanie
-
‘Attack-resistant Construction of Rooted Spanning Tree Embeddings in Friend-to-Friend Overlays’, TU Dresden
Martin Byrenheid
-
PolySphinx: Extending the Sphinx Mix Format With Better Multicast Support. 2024 IEEE Symposium on Security and Privacy (SP), 4386-4404.
Schadt, Daniel; Coijanovic, Christoph; Weis, Christiane & Strufe, Thorsten
