A Multiagent Learning Approach for Distributed Control of Address Randomization in Communication Destination Anonymization

Keeping anonymity of communication destination in networking is one of the important issues to be improved since sniffing packets can still be a major threat especially on a local network system. In 2017, U-TRI has been proposed by Wang et al. as one of the approaches to provide better anonymity in such a context with acceptable overheads. However, as they mentioned, U-TRI still suffers from the issues that allow attackers to utilize their observed traffic trends. In this paper, we present an approach to solve this issue by introducing a multi-agent learning for autonomously coordinating multiple end-hosts and a simulation environment to analyze it.