Ask HN: Can you make a voting/rating system without logins?
Posterous has done well eliminating the login hurdle.
Has anyone been able to do this for a voting/rating site?
I've been thinking about this for my site BuyersVote.com. I let people do some actions (like start a new category) without logging in, but rating I haven't figured out how to avoid abuse without logins. Offering logins with OpenID (and rpxnow.com) reduces the hurdle significantly since people don't have to create a new account, they can use an existing one, but it's a hurdle none the less. Thanks for any tips!
(
38 comments
| 19 points
| on HN
)
...no you can't make a credible voting or rating system without a login.
It's too easy to game the system to create repeat votes via cookie burning, ip rotation, etc (see the other responses).
Forcing someone to be logged in and then only allowing one vote per poll helps keep all but the most persistent griefers.
To that end, be aware account creation can be automated and unlimited unique email addresses created from one domain.
However, the difference between gaming the system by creating lots of fake accounts and gaming the system via cookie destroying/IP rotation is that the former is easy noticed within your system - assuming you have built such checks in place. It's also lets you easily remove the offending over-votes to restore calm to your survey/poll/etc.
The fingerprint does not eliminate every attempt of cheating, but the method does not rely on cookies and just changing your ip-adress by reconnecting is also not helping, because you use the geolocation and not the IP. TOR, or just changing Proxies would still work, as well as using different browsers.
Another method I can think of are flash cookies. Most users don't know about them. Let users vote as often as they like, but only count them once. In order to make sure they believe their votes count you can display the result personally for them and include their additional votes, so they won't go looking, how they are recognized.
And if a user does not know about flashcookies it's pretty much impossible to cheat the system - and the flashcookie remains the same regardless of the browser you use.
This has all the flaws that other people here have pointed out (multiple users behind same IP etc), but I feel that the advantage of being easy to use is greater than these disadvantages. It serves it main purpose (most of the time), of putting good content on top (see e.g. http://tagdef.com/ff) and hiding bad content.
For the JavaScript fingerprinting, I wonder if there are ways to probe for bugs that characterize different browser versions both for a richer fingerprint and to detect spoofing of the useragent.
Other things that can fit into a credibility score: B computational JavaScript payload to impose more costs on automated abusers. Elapsed time between when the voting form was returned, and when the vote was submitted.
On the IP front, keep in mind the ability to correlate not just IP, but also netblock.
Sadly, it's extremely easy to game it. See http://news.ycombinator.com/item?id=913966
Btw, I have not tested this. Use this at your own risk.
At least theoretically speaking, if we can model and predict human behavior, then the best way to do it is to let humans behave as naturally as possible, and model a rating system based on that behavior.
1. Fire up Chromium in incognito, and once you've voted once, just close it again and open it again to vote once more. Rinse and repeat.
2. The ip choice is smart, but there are ways to get a dynamic ip every [insert time interval].
One could perfect this to an art really. Get your ip to start changing, have Chromium open on a hotkey, use ctrl+shift+n to go incognito (or just have it auto-load in incognito), have the poll set as your homepage. You can clock alot of votes this way.
Logins are best.
A couple years ago, some university friends and I went to an engineering competition in Ontario. It was an event spread over four or five days, and on the last night a dance was held. This information was given to the participants a couple of weeks in advance.
The DJ for the dance had set up a website to allow people to request songs from off of a giant list. Each person got five or six "votes", and the system stopped them from voting any more after that. It was a cookie-based system. (It didn't bother with IP stuff).
One of my friends ended up using the macro system on his Macbook to vote for the song he wanted until he ran out of votes, delete his cookies, refresh the page, and repeat. He ran this macro overnight for a couple of days.
By the time we got to the competition, "Never Going to Give You Up" had three or four thousand votes. There were about two hundred people attending.
If someone's willing to go to that far of an extreme to game a vote, I don't think a login system would do much to stop them either. You can just as easily create new accounts with the new IP and a fake email address.
The difference is that with a login method, it takes significantly more effort and time. This is pretty straightforward once you have it setup.
Is there ever a situation where a bunch of legitimate users have the same IP? I'm thinking at a university maybe, corporate firewall, or AOL?
Even so this compromise might be worth it to remove the login hurdle.
If you become Digg or something and people start actively trying to get around it, do you think this would be about as secure as logins? Could be a high quality problem to have.