paint-brush
Impact of Packet Loss on OpenVPN Detection Accuracy and Censoringby@virtualmachine
New Story

Impact of Packet Loss on OpenVPN Detection Accuracy and Censoring

by Virtual Machine TechJanuary 13th, 2025
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

This research shows packet loss over 10% reduces OpenVPN detection accuracy, leading to underblocking, which aligns with censoring priorities.
featured image - Impact of Packet Loss on OpenVPN Detection Accuracy and Censoring
Virtual Machine Tech HackerNoon profile picture
0-item

Authors:

(1) Diwen Xue, University of Michigan;

(2) Reethika Ramesh, University of Michigan;

(3) Arham Jain, University of Michigan;

(4) Arham Jain, Merit Network, Inc.;

(5) J. Alex Halderman, University of Michigan;

(6) Jedidiah R. Crandall, Arizona State University/Breakpointing Bad;

(7) Roya Ensaf, University of Michigan.

Abstract and 1 Introduction

2 Background & Related Work

3 Challenges in Real-world VPN Detection

4 Adversary Model and Deployment

5 Ethics, Privacy, and Responsible Disclosure

6 Identifying Fingerprintable Features and 6.1 Opcode-based Fingerprinting

6.2 ACK-based Fingerprinting

6.3 Active Server Fingerprinting

6.4 Constructing Filters and Probers

7 Fine-tuning for Deployment and 7.1 ACK Fingerprint Thresholds

7.2 Choice of Observation Window N

7.3 Effects of Packet Loss

7.4 Server Churn for Asynchronous Probing

7.5 Probe UDP and Obfuscated OpenVPN Servers

8 Real-world Deployment Setup

9 Evaluation & Findings and 9.1 Results for control VPN flows

9.2 Results for all flows

10 Discussion and Mitigations

11 Conclusion

12 Acknowledgement and References

Appendix

7.3 Effects of Packet Loss

We investigate the effects of packet loss on the performance of the Filter. An adversary analyzing traffic on a busy network needs to keep pace with the packet arrival rate, or otherwise packet drops will start to occur due to a CPU bottleneck. For the opcode and ACK fingerprinting, we need to inspect the raw contents of each reassembled packet until the observation window is reached, for all flows. This means all traffic must be passed on to Zeek’s scripting layer, which may lead to a CPU bottleneck. In addition, the Network Interface Card (NIC) may also become an upstream bottleneck and lead to end-to-end packet loss. We therefore explore what to expect from the Filter when packet loss is inevitable.


We configure Zeek to ignore events signaling new packets with different probabilities in order to simulate random packet loss. We test loss rates from 1% up to 80%. The experiment is repeated three times, and the average result is reported in Figure 8 (b). We find that packet loss starts to have noticeable effects on the Filter’s outputs once the loss rate surpasses 10%. Notably, when packet loss starts to affect the detection


Figure 8: Effects of (a) observation window and (b) average packet loss rate on Filter’s performance


accuracy, both opcode and ACK fingerprint vulnerabilities always produce underblocking instead of overblocking, which is favoured by real-world censors [56]. Still, in order to minimize the effects of packet loss, we always configure the Monitoring Station to sample flows with a given rate (adjusted with CPU resources and traffic volume) for our online evaluation.


This paper is available on arxiv under CC BY 4.0 DEED license.