The researchers were led by graduate student Albert Kwon. In a paper published online earlier this month, the group detailed how Riffle could emerge as a next-generation communication system with strong anonymity. It enables more efficient exchange of messages than is possible with Tor, increasing performance.
Tor has become the default program to use when communicating securely online. It has gained popularity with a wide selection of demographics, varying from people who want to protect their personal privacy to criminals trading stolen and illegal goods.
While Tor has largely proved to be resistant to external attacks, it contains weaknesses that one day could result in the destruction of the network. Tor works by using a system of relay nodes to bounce traffic between computers across the world, anonymising the data so it’s impossible to trace the original source.
The issue is that attackers could start to work out where traffic comes from if they came to control several of the nodes. With a substantial portion of the network under their control, they could theoretically monitor the entirety of Tor and piece together a trail linking a user to individual packets of data.
Riffle operates in a different way. According to its creators, it provides greater security and performance in the process. The anonymization begins by continually switching around the order in which servers pass on data to each other, preventing the construction of a breadcrumb trail or analysis of packets to work out where data has come from.
Afterwards, a two-step procedure ensures infected servers on the network are unable to intercept packets and replace them with tracking ones. Data transmitted by a user is distributed to every active Riffle server. When a server passes a packet on, it is required to sign it with a digital signature, based on a complex mathematical proof that can be externally verified.
This prevents any malicious servers on the network from intercepting data being transmitted using Riffle. Tor begins anonymization by directing user data to just one server. If that server is operated by attackers, the data could be hijacked or tampered with before it reaches its destination.
The two techniques used by Riffle aren’t anything new. In the past, they’ve been too complex to implement though, laden with issues that prevent them from being put to use. A major caveat is the high amount of bandwidth required and the CPU time needed to generate mathematical proofs. The result would be a network significantly slower than Tor.
The MIT researchers made a breakthrough that allows them to overcome these problems. By employing cryptographic techniques commonly used by web services and apps, Riffle uses a system of public keys that allows the issues to be circumvented. The network is much better protected against attack than Tor is but doesn’t suffer from latency or bandwidth issues.
In testing, MIT saw Tor complete file transfer operations in one tenth the time needed by Tor. It achieved bandwidth of over 100KB/s per user for 200 users and could handle over 100,000 users with a latency of less than 10 seconds while microblogging.
Despite the benefits, Riffle is intended to be a prototype rather than a true alternative to Tor. It works best with a relatively small number of servers as the network actually slows down as nodes are added. Sending every packet of data to every server ends up being a costly operation if hundreds of thousands of nodes are online, since the original sender must wait for them all to finish processing.
There’s currently no plans to publicly release Riffle as an alternative to Tor. Kwon does intend to launch a downloadable version in the future though, after a code tidy-up has been completed. Riffle demonstrates that Tor isn’t the most secure network possible, providing a glimpse into the future of online anonymity.
