LiveJournal's method for blocking spam

LiveJournal is an online community of blogs. They have recently implemented a new way of keeping out spammers on blog comments. We recently discussed the MovableType Plugin that will allow MovableType (perhaps TypePad.com users as well?) users to block spammed comments.

The way these comments come about is that a bot is written to scan the blogs for keywords, and then when it finds them, it replies to the message. Usually it says something rather vague like "I read this." and then in the message is a link to a website. Usually a porn site or a casino.
The reason these bots are written is because of Google (and to some much smaller extent, so that readers of blogs will see and click on these links as well). Google's algorithm for figuring out what makes a good page is partly influenced by how many other pages link to it. So if it scans blogs and finds that a lot of comment pages link to a site, it will then rank that site higher. So even though the comments are useless, the automated Google system doesn't know this.

LiveJournal has implemented a system that watches an IP address that is posting to the LJ system. If it posts a certain number of comments anonymously (it might have name info and such so that it doesn't look anonymous - but they mean that it isn't a member of the LJ community and/or not logged in), then after some set number of posts in a short time period, it will trigger a system where it will require a test to post more.
That test is one where it has to read an image and put what the image contains (usually alphanumeric text) into a textbox as part of the submission process.

This means that you might get 10 comments in quick succession from a bot, but after that it will be blocked out. You can then easily deal with the 10 comments (delete them) and be done with it.
If it happens to be a real person just posting really quickly, then they can still do so by looking at the graphic and entering the data.

That isn't to say that there aren't still flaws in this system.
1) it assumes that all spammers are going to be anonymous. This is a fairly good assumption actually since you have to pay for the LJ system (unless you are grandfathered in from when there was a free system). Spammers are not likely to pay for a system to spam in where they will likely get banned shortly after they start - it is cheaper/easier for them to focus on free systems where the only cost is the effort.

2) it assumes that all real people that happen to be posting a lot can pass the test to get through. It has been shown that visually impaired people don't usually do well on those tests - since usually they are using a computer to read the text on screen to them in the first place - if you then use a system that is designed to baffle a computer, it is going to prevent them from posting as well.
But I have a feeling the demographic of LiveJournal posters that quickly write many comments and also happen to be visually impaired is a small enough one that LJ can essentially ignore that issue.

So all in all, this is one way to slow down the spammers on the LJ system. It doesn't totally stop them - there are still the few spams that get through before it triggers the system - but it is better than letting them run totally unchecked.

Posted by Eric at March 11, 2004 10:23 AM | TrackBack