The Tor network uses a measurement system called TorFlow to estimate its relays' forwarding capacity and to balance traffic among them. This system has been shown to be vulnerable to adversarial manipulation, and inaccuracies even in benign circumstances have long been observed. To solve the issues with security and accuracy, we present FlashFlow, a system to measure the capacity of Tor relays. Our analysis shows that FlashFlow limits a malicious relay to obtaining a capacity estimate at most 1.33 times its true capacity. Through realistic Internet experiments, we find that FlashFlow measures relay capacity with$\geq {89\%}$accuracy 95 % of the time. Through simulation, we find that FlashFlow can measure the entire Tor network in less than 5 hours using 3 measurers with 1 Gbit/s of bandwidth each. Performance simulations using FlashFlow for load balancing shows that, compared to TorFlow, network weight error decreases by 86 %, while the median of 50 KiB, 1 MiB, and 5 MiB transfer times decreases by 15 %, 29 %, and 37 %, respectively. Moreover, FlashFlow yields more consistent client performance: the median rate of transfer timeouts decreases by 100 %, while the standard deviation of 50 KiB, 1 MiB, and 5 MiB transfer times decreases by 55%, 61 %, and 41 %, respectively. We also find that the performance improvements increase relative to TorFlow as the total client-traffic load increases, demonstrating that FlashFlow is better suited to supporting network growth.

Tor’s growing popularity and user diversity has resulted in network performance problems that are not well understood, though performance is understood to be a significant factor in Tor’s security. A large body of work has attempted to solve performance problems without a complete understanding of where congestion occurs in Tor. In this article, we first study congestion in Tor at individual relays as well as along the entire end-to-end Tor path and find that congestion occurs almost exclusively in egress kernel socket buffers. We then analyze Tor’s socket interactions and discover two major contributors to Tor’s congestion: Tor writes sockets sequentially, and Tor writes as much as possible to each socket. To improve Tor’s performance, we design, implement, and test KIST: a new socket management algorithm that uses real-time kernel information to dynamically compute the amount to write to each socket while considering all circuits of all writable sockets when scheduling cells. We find that, in the medians, KIST reduces circuit congestion by more than 30%, reduces network latency by 18%, and increases network throughput by nearly 10%. We also find that client and relay performance with KIST improves as more relays deploy it and as network load and packet loss rates increase. We analyze the security of KIST and find an acceptable performance and security tradeoff, as it does not significantly affect the outcome of well-known latency, throughput, and traffic correlation attacks. KIST has been merged and configured as the default socket scheduling algorithm in Tor version 0.3.2.1-alpha (released September 18, 2017) and became stable in Tor version 0.3.2.9 (released January 9, 2018). While our focus is Tor, our techniques and observations should help analyze and improve overlay and application performance, both for security applications and in general.

Experimentation tools facilitate exploration of Tor performance and security research problems and allow researchers to safely and privately conduct Tor experiments without risking harm to real Tor users. However, researchers using these tools configure them to generate network traffic based on simplifying assumptions and outdated measurements and without understanding the efficacy of their configuration choices. In this work, we design a novel technique for dynamically learning Tor network traffic models using hidden Markov modeling and privacy-preserving measurement techniques. We conduct a safe but detailed measurement study of Tor using 17 relays (~2% of Tor bandwidth) over the course of 6 months, measuring general statistics and models that can be used to generate a sequence of streams and packets. We show how our measurement results and traffic models can be used to generate traffic flows in private Tor networks and how our models are more realistic than standard and alternative network traffic generation~methods.