PriFi

PriFi [1] is an anonymous communication network with low-latency and traffic-analysis resistance, co-developed at EPFL and Yale University. It provides protection against eavesdropping attacks and against traffic analysis on organizational networks, for instance perpetrated by rogue and coerced employees, malware, or via parking-lot attacks.

Read the paper : https://petsymposium.org/2020/files/papers/issue4/popets-2020-0059.pdf

Get PriFi : github.com/dedis/prifi

PriFi in 2 minutes

Please see here for the 15-min presentation delivered at PETS 2020.

What PriFi provides

How it works

PriFi is built upon Dining Cryptographer networks [3], a cryptographic primitive that provides provable anonymity.1 It is tailored for WLANs and LANs, hence it works best at your company, on your university campus, or in your home. PriFi is like a VPN, but does not need to be trusted. Install it on your users computers, and it transparently anonymizes any kind of traffic.

PriFi uses a client-server infrastructure for performance and security. It relies mostly on existing infrastructure : a relay/router, a set of clients, and some additional public servers, possibly shared among several organizations.

PriFi architecture
Figure 1 : PriFi architecture relies on existing relay/clients, and requires additional public servers.

These servers provide security; the first interesting property is their security model, called anytrust [4]. It means that as long as any one of the selected servers is honest, PriFi will keep its security guarantees. In practice, your organization can select a set of servers it trusts; these servers can become compromized, but as long as not all of them are compromized, PriFi will deliver strong anonymity.

The second interesting property is the path taken by the data: Unlike Tor and other mixnets, the anonymized data does not go through the servers. This is important because the latency to these servers is usually orders of magnitude above the latency in a WLAN/LAN; In PriFi, this high-latency path matters only at the setup. Once the setup phase is complete, the packets from the clients to the Internet follow their usual path, with no added hop that would increase latency.

PriFi high- and low- latency paths
Figure 2 : Low-latency path (order of magnitude : 10ms) and high-latency path (order of magnitude : 100ms).

1 This construction provides perfect anonymity; in particular, it is resistant to traffic-analysis attacks [5], unlike Tor (and other systems). These attacks exploit the differences in traffic flows among users, and these differences are an effective way to de-anonymize users.

People

Bryan Ford

Prof. Bryan Ford
EPFL

www

Joan Feigenbaum

Prof. Joan Feigenbaum
Yale

www

Jean-Pierre Hubaux

Prof. Jean-Pierre Hubaux
EPFL

www

Mahdi Zamani

Mahdi Zamani
Yale

www

Italo Dacosta

Italo Dacosta
EPFL

www

Ludovic Barman

Ludovic Barman
EPFL

www

Ennan Zhai

Ennan Zhai
Yale

Apostolos Pyrgelis

Apostolos Pyrgelis
EPFL

References

  1. [1] Barman, Ludovic; Zamani, Mahdi; Dacosta, Italo; Feigenbaum, Joan; Ford, Bryan; Hubaux, Jean-Pierre and Wolinsky, David. PriFi: A Low-Latency and Tracking-Resistant Protocol for Local-Area Anonymous Communication. Proceedings of the 2016 ACM on Workshop on Privacy in the Electronic Society
  2. [2] Dingledine, Roger; Mathewson, Nick and Syverson, Paul. Tor: The second-generation onion router.
  3. [3] Chaum, David. The dining cryptographers problem: Unconditional sender and recipient untraceability. Journal of cryptology
  4. [4] Wolinsky, David I; Corrigan-Gibbs, Henry; Ford, Bryan and Johnson, Aaron. Scalable anonymous group communication in the anytrust model.
  5. [5] Raymond, Jean-François. Traffic analysis: Protocols, attacks, design issues, and open problems. Springer
  6. [6] DeDiS Lab, EPFL. Secure Distributed Algorithm. https://github.com/dedis/cothority

Cette page appartient aux archives web de l'EPFL et n'est plus tenue à jour.

This page belongs to EPFL's web archive and is no longer updated.