Comment by fiaz | Hacker News Reader

Comment by fiaz | Hacker News Reader

The simplest technique would be for the YC News gods to grant us a "report SPAM" button. The site would then be self moderating.

I would also suggest something like granting karma points for sniffing out SPAM in case the YC News gods agree that a reported link does indeed turn out to be spam.

Alternatively, a system of "open moderation" would be good as well (to lessen PG's load). In this case, moderation of the site would fall upon a rotating schedule between users that have reached a certain level of karma points. Like jury duty, you have a queue of sites that have been flagged for review, and the member's job is to merely approve of something as spam or not and have it dealt with. A link would have to be scrutinized by the random "committee" (which ideally should be a random percentage of YC News members on a regular schedule) and a majority vote would determine collectively if a link is spam or not.

---------------

Apologies for making this post so long....

1) have a "report" button

2) who ever is on "spam patrol" would see a reported link highlighted

3) "spam patrol" would have their "report" button substituted by a "confirm" button

4) after enough of the "spam patrol" confirms a link as spam, then the link is automatically marked for deletion

The advantage of such a system is that there would be some filtration of the links and spammers would (hopefully) be discouraged.

An "emergent" advantage would be to uncover spam sites. I can even imagine a number of social sites adopting a scheme creating a collaborative spam filter (ok, yes I'm spinning this out of control, but what the hell, I'm sure everybody here thinks like this!). This would be a social solution to a technical problem.

If reddit had such a system, then a collaboration between YC News and reddit would pre-emptively eliminate confirmed spam sites from one another. Using this example, a number of social sites could immediately benefit from an open database of spam filtering (akin to the use of OpenID). So a confirmed spam site on reddit would automatically translate into a confirmed spam site on YC News, or whoever else subscribes to such an open database.

Is this too complex of an idea???

It could be tricky to get right. For example, someone could censor a political blog from a social site interested in politics by spamming it to sites that are not interested in politics.

And, of course, if this kind of filtering gets popular, spammers can simply set up a separate website for each submission to a social site.

Awesome point about the political blog example. Perhaps attaching some sort of meaning to the filtering instead of a black/white SPAM vs. NOT-SPAM would be an added advantage of leveraging human judgement. So for example spam on a political oriented site would include news about celebs; or spam on gawker.com would equate to political links.

So the levels of filtration could be:

1) SPAM

2) Not relevant (to celeb/political/geek/etc nature of site)

3) <suggestions please.../>

The problem with this is the notion of "purity" between a collection of social news sites. Too much homegeneity (is that a word?) is problematic as diversity is what brings value to social sites. I suppose having a level of how stringent the filtration should be an option for subscribing to such a database. There should also be a way to decentralize this as centralization is not good (just my opinion...).

As far as spammers setting up separate websites for each social site....GOOD POINT!! A Bayesian filter that would introspect different links in a database could help match different sites at the content level with some degree of confidence, thus exposing the level of sophistication a specific spammer and identifying higher spam threats. Similar content could be matched between different databases (if decentralized) and such information could propagate faster between different types of social sites identifying spammers by content.

Because human judgment would trickle down to a central location (how can this be decentralized I wonder...) means that the timing of the appearance of sites in the database would possibly be significant (wouldn't it?), adding yet another insight into how spammers might operate (ok, perhaps I'm stretching it here).

I think the best way to look at it is to have a human filter (that depends on mass collaboration) up front and then later use a Bayesian filter to associate at the content level the different websites. Creating a web crawler to do this would be pretty trivial (I would imagine).

------------------

How would spammers game such a system? One thing they can do is NOT report their own links as spam. This does not prevent others from doing so. Spammers could conversely report other sites as spam thus muddying the waters through volume. I can imagine a scenario where they would muddy the waters through their own sites, which would translate into merely blocking that site as a preventive measure (this is DEFINITELY fraught with problems that I have not thought of yet!!!). Similarly if they attempt to muddy the waters through a legit site, then there is the advantage of genuine human involvement and randomness.

This becomes an issue of membership to the database: what are the criteria to submit to the database and be a contributor to flagging spam sites? I suppose subscribing to such a database would be free for all.

A possible solution to centralized authority would be for everybody to contribute, but at the same time a site can allow to trust the quality of reporting from certain sites and exclude others; this could have yet another "emergent" quality of identifying collaborations of sites that wall off spammers trying to muddy the water: spammers can muddy the water as much as they want, but they won't be heard. This would eliminate the need described above for judging different sites; also, it could provide new sites for a basis of which sites to trust for authority. Theoretically the head of a long tail of trust should emerge, and hopefully this could possibly be a shifting mass of authority. This decentralization of authority would allow spammers to do their worst and still not have an effect (wouldn't it?).

But somehow I feel intuitively there might be a way to game the system...any thoughts???

Another easy way for spammers to game the system: Set up a link with a copy of a legitimate, interesting news from another location (perhaps from a social site with similar interests) and submit it to a social site. Wait a few hours while the users classify it as non-spam and comments start to roll in. Then, change the page to be mostly spam. (You can't fight this by taking hashes or diffing pages; most pages these days have dynamic content.)

Not only does this spread the spammer's message, but it also breaks down your proposed web of trust between social sites; sites that approve links that later turn out to be spam will be seen as untrustworthy.

Yes, this would break it temporarily. In the long run this strategy would lose. With enough eyeballs scrutinizing a site, such a system would eventually identify a site as spam and have it marked because of the dynamic nature of human judgement.

So let's say that out of 100 visitors, 20 see legitimate content, the next 20 see spam, thus alerting the system that they have encountered a spam site. The Bayesian filtering would be for sites with established spam content and then matching it against other established spam content. This would be a cost of 20 visitors exposed to spam while 20 others are exposed to legitimate content. The remaining 60 users are diverted. Sounds like a reasonable tradeoff to me...

Spammers could keep on shifting the content of sites, and this would make it more difficult to apply the spam filter for similar content (but humans, the first line of defense, aren't so easily fooled); this is where site moderators can identify users as potential spammers and give them the boot. So instead of having a 1:1 filter for domain:content, you could have a 1:n filter and then match that against other sites with similar 1:n "content shifting". A danger here is that legitimate content gets associated with spam and thus gets filtered out; this would require more sophistication on behalf of the Bayesian filter in associating content - but I'm certain this is a solved problem that can be put to use here.

It's more complex than: add CAPTCHA to create account screen.

You are right; at the same time there is nothing novel about CAPTCHAs. I also have a habit of running wild with ideas, and I suppose it's beginning to show...

But hopefully my hatred for spam is showing through as well!

Not all spam is automated.

I run another social news site (realestatevoices.com) and captcha doesn't help. The people inputing these links are 'SEO' firms based out of India.

Validating email accounts is worthwhile though. But keep in mind these spammers are actual people not automated in many cases.

interesting site...how long have you been up and running?

Reply on news.ycombinator.com