Friday, November 03, 2006

Tamper-proof voting systems

Digg is a collaborative news site, much like Slashdot, where people post articles that seem interesting and topical. Through a voting scheme, articles get "dugg", and rise to the top of the site. In fact, getting "dugg" is now an inadvertent denial-of-service attack much like being "slashdotted" used to be.

The problem is that people attempt to game the system to bump up traffic to their sites, by forming voting coalitions, making fake accounts, etc etc. Needless to say, there can often be real money (in the form of advertising) behind this, so there's a lot of incentive to cheat the system.

This is not a new problem; Google and other search engines have battled search engine optimizers for a long time now. There are companies that claim to improve your location in a Google search result (for a small fee of course). An interesting side effect of all this gamesmanship is that search engines like Google and aggregators like Digg have to keep many details of their ranking process secret. Obviously, some of this is for IP protection, but protecting against vote riggers is also a major concern. A tertiary consequence of this lack of transparency is that such sites are now vulnerable to lawsuits by disgruntled sites complaining about bias in search results; Google has had to fend off lawsuits based on some claim of bias leading to monetary damage.

The latest such problem has hit Digg, where in an attempt to block out users trying to game votes on articles they want to push, the management has managed to freeze out and frustrate some of the most prolific users. A user-driven site like Digg that has many competitors can't really afford to be annoying its most valuable contributors.

So (finally), here's the technical question. Although Arrow's theorem tells us that in general, voting schemes can always be defeated, I don't know if the result is constructive. In other words, even if there is a voting strategy that can break one of the criteria for a reasonable voting scheme, it may not be easy to find such a scheme.

So, in the spirit of RSA, is there a way of designing a voting scheme that can be published (thus addressing issues of transparency), but is computationally intractable to game ? Any cryptographers know if this has been studied ?


Disqus for The Geomblog