Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Decentralized Reliability Estimation for Low Latency Mixnets

Claudia Diaz, Harry Halpin, Aggelos Kiayias

TL;DR

This paper tackles the challenge of achieving verifiable reliability in low-latency mixnets without sacrificing performance. It introduces a decentralized reliability estimation framework that uses indistinguishable measurement packets and a novel VRF-based routing primitive to produce publicly verifiable scores for links and nodes, with estimation overhead decoupled from total traffic and limited latency impact. Reliability is inferred from epoch-based measurement data, leveraging Beta/Bernoulli statistics to bound errors, and is augmented by a threshold-based attribution mechanism to handle adversarial behavior. Experimentally, the authors demonstrate accurate edge- and node-level reliability estimates in both unreliable and adversarial settings, and provide a detailed protocol overhead analysis showing practical feasibility for real-world deployments like Nym. The work offers a scalable path toward accountable, low-latency mixnets suitable for broad applications while preserving privacy and avoiding single points of failure in reliability monitoring.

Abstract

While there exist mixnets that can anonymously route large amounts of data packets with end to end latency that can be as low as a second, %making them attractive for a variety of applications, combining this level of performance with strong verifiability and reliability properties that ensure the correct processing and delivery of packets has proved challenging. Indeed, existing verifiability mechanisms are incompatible with scalable low-latency operation due to imposing significant latency overheads measuring in minutes to hours, hence severely limiting the variety of applications mixnets can serve. We address this important gap by proposing a scheme that can estimate reliability scores for a mixnet's links and nodes in a decentralized manner with essentially optimal complexity that is independent of the total traffic routed through the mixnet. The scores can be computed publicly by all participants from a set of measurement packets that are eventually revealed and act as a random sample of the traffic, without affecting mixnet transmission latency for client packets or incurring significant bandwidth overhead. Our scheme assumes client credentials and relies on VRF-based routing, a novel primitive that ensures that legitimate client packets follow the routing policy of the mixnet, as well as randomly generating unforgeable measurement packets. We experimentally validate our construction both in unreliable and adversarial settings, demonstrating its feasibility.

Decentralized Reliability Estimation for Low Latency Mixnets

TL;DR

This paper tackles the challenge of achieving verifiable reliability in low-latency mixnets without sacrificing performance. It introduces a decentralized reliability estimation framework that uses indistinguishable measurement packets and a novel VRF-based routing primitive to produce publicly verifiable scores for links and nodes, with estimation overhead decoupled from total traffic and limited latency impact. Reliability is inferred from epoch-based measurement data, leveraging Beta/Bernoulli statistics to bound errors, and is augmented by a threshold-based attribution mechanism to handle adversarial behavior. Experimentally, the authors demonstrate accurate edge- and node-level reliability estimates in both unreliable and adversarial settings, and provide a detailed protocol overhead analysis showing practical feasibility for real-world deployments like Nym. The work offers a scalable path toward accountable, low-latency mixnets suitable for broad applications while preserving privacy and avoiding single points of failure in reliability monitoring.

Abstract

While there exist mixnets that can anonymously route large amounts of data packets with end to end latency that can be as low as a second, %making them attractive for a variety of applications, combining this level of performance with strong verifiability and reliability properties that ensure the correct processing and delivery of packets has proved challenging. Indeed, existing verifiability mechanisms are incompatible with scalable low-latency operation due to imposing significant latency overheads measuring in minutes to hours, hence severely limiting the variety of applications mixnets can serve. We address this important gap by proposing a scheme that can estimate reliability scores for a mixnet's links and nodes in a decentralized manner with essentially optimal complexity that is independent of the total traffic routed through the mixnet. The scores can be computed publicly by all participants from a set of measurement packets that are eventually revealed and act as a random sample of the traffic, without affecting mixnet transmission latency for client packets or incurring significant bandwidth overhead. Our scheme assumes client credentials and relies on VRF-based routing, a novel primitive that ensures that legitimate client packets follow the routing policy of the mixnet, as well as randomly generating unforgeable measurement packets. We experimentally validate our construction both in unreliable and adversarial settings, demonstrating its feasibility.
Paper Structure (29 sections, 1 theorem, 7 equations, 4 figures, 2 tables)

This paper contains 29 sections, 1 theorem, 7 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Model and Problem Statement
  3. System Model
  4. Threat Model
  5. Properties
  6. Reliability Estimation
  7. Link reliability estimation
  8. VRF-Based Routing
  9. Node reliability estimation
  10. Simulation-based empirical evaluation
  11. Unreliable setting
  12. Adversarial setting
  13. Protocol overhead evaluation
  14. Related Work
  15. Conclusion
  16. ...and 14 more sections

Key Result

Theorem 1

Assuming the pseudorandomness of the VRF, the Decisional Diffie Hellman assumption and that $H(\cdot)$ is modeled as a random oracle. (I) any security or privacy attack against the VRF-based encoding presented above can be transformed to an attack against the underlying mixnet encoding scheme with n

Figures (4)

  • Figure 1: Illustrating the dependencies between the packet variables $\alpha, b, s, r, t, \tilde{r}$ when processing a VRF-based routing packet in the case of $\nu=2$. The measurement packet condition is illustrated in the upper left.
  • Figure 2: Distribution of error ${\epsilon}_j = \hat{\rho}_j - {\rho}_j$ for unreliable (blue) and reliable (black) nodes, for different amounts of measurement samples (from $25$k to $2$ million per epoch). We run $20$ simulations per setup, each resulting in $320$ values of ${\epsilon}_j$ (i.e., $6400$ samples per pair of boxplots). Each boxplot shows the median (orange line), the first and third quartiles (upper and lower limits of the box), and the range of the distribution (whiskers); outliers are not depicted.
  • Figure 3: Results for adversarial settings where all honest nodes are reliable. Each sample represents the aggregate reliability penalties $c_T$ for $1\leq |T|\leq 64$ targets ($y$ axis) and $c_A$ for $1\leq |A|\leq 64$ adversaries ($x$ axis) in each simulation run (total $510$ runs).
  • Figure 4: Each sample represents the aggregate reliability penalties $c_T$ for $1 \leq |T| \leq 64$ targets ($y$ axis) and $c_A$ for $1 \leq |A| \leq 64$ adversaries ($x$ axis) in $630$ simulation runs. Purple circles represent scenarios where adversaries control at most $20\%$ ($16$ out of $80$ nodes) of any layer. Green '+' signs correspond to simulations where adversaries make up $40\%$ ($32$ out of $80$ nodes) of at least one layer.

Theorems & Definitions (3)

  • Definition 1
  • Theorem 1
  • proof