Master Thesis
Peer Discovery in the Presence of Rational Adversaries
In decentralized file-sharing systems, users possess files they wish to exchange with others who hold complementary resources. Efficient peer discovery is crucial for enabling these exchanges, but traditional protocols assume honest participation. In reality, rational adversaries may strategically withhold information about other peers to maximize their own utility, creating information asymmetries that can severely degrade system performance or even partition the network.
This thesis explores peer discovery protocols in incomplete but connected network graphs where participants act as rational agents rather than purely honest or malicious entities. Unlike complete graphs where every peer knows every other peer, realistic topologies require users to rely on intermediaries for discovery. However, these intermediaries have incentives to manipulate the discovery process: for example, by concealing the existence of certain peers to maintain exclusive trading relationships or to extract higher value from their own exchanges.
The student should formalize threat models for rational adversaries with various utility functions. The thesis should then investigate how different graph structures, such as random graphs, scale-free networks, small-world networks, and structured overlays, affect the adversary’s ability to manipulate discovery. The analysis should characterize conditions under which honest peer discovery can be guaranteed or incentivized, and explore mechanism design approaches (game-theoretic solutions, reputation systems, or cryptographic commitments) that align individual incentives with global discovery efficiency.
References
[1] Revisiting Rational Broadcast Protocols
[2] Incentives Build Robustness in BitTorrent