Ask HN: Review my POC to detect spam accounts on twitter | Hacker News Reader

Ask HN: Review my POC to detect spam accounts on twitter

6 pointssucuri27 comments

Hey HN,

I have been complaining a lot lately about spam on twitter and decided to to something.

So far it is just a POC (proof of concept), where you provide the twitter account name and it will give a score from 0 to 10 and let you know if the account is spammer/bot or not.

Link: http://sucuri.net/?page=docs&title=twitter

How it works? We use a point-based system and consider some of the things:

-Account name (looks spam?)

-Profile image (default?)

-Bio (valid, etc?)

-URL (valid, pointing to a valid domain)

-updates (how many)

-How many links per update

-Repeated links

-Spam words

-How many interaction with other users

-Number of followers/following

And a few more... Please report FP if you found them.

Later it can be integrated with other tools (or even search tools) to avoid following spammers (or even to auto-block them).

Thanks!

I think this is a fantastic idea, but the filter could stand to benefit from better requirements. For example number of direct messages sent to the user (since people don't DM spam bots very often), number of @replies sent to the user.

It could stand to happen that a lot of users who use sites like Brightkite or other sites where they broadcast what they're listening to or where they're at could get caught u p in your spam/bot filter simply for making heavier use of such sites.

I'd revisit your filter to look at how the user interacts with other users, versus just monitoring posting behavior.

The problem I think is that DMs are not public information... You can only know yours, not from the others.

Yeah, you're right, and I knew that too. Heh. Thanks.

You made some very good points. But we also take into consideration the number of followers, so if sites like slashdot or Brightkite broadcast their info and have lots of followers, the spam count will go down too.

Lots of variables in there and I plan to improve even more.

This is the direct link:

http://sucuri.net/?page=docs&title=twitter

Good one.

What about a user driven blacklist?. Maybe you can collect them the way @spam (http://twitter.com/spam) does.

That's a good idea too. Will see how this can be integrated with them.

Reply on news.ycombinator.com