Improving Kademlia Lookup Privacy through Query Obfuscation
Erik Daniel, Guillaume Michel, Florian Tschorsch · 2025
This paper addresses privacy challenges inherent in Distributed Hash Tables (DHTs). While DHTs facilitate efficient content lookup, privacy concerns arise due to query mechanisms revealing user interests. In the paper, we focus on Kademlia-based DHTs and propose to obfuscate the lookup item by presenting three obfuscation methods: double hashing, Private Set Intersection, and prefix fetching. Based on our privacy improvements, we present a protocol specification for the libp2p kad-dht, a popular Kademlia implementation. The methods are analyzed in the context of measurement values derived from the public IPFS network, which uses kad-dht.