more thoughts on LJ spam

[charmian]

LJ has released a further statement on the spam issue.

http://news.livejournal.com/135748.html

In any event, azurelunatic had an idea about reporting spam on the "Latest LJ posts" page. I have no idea whether this will work (someone I talked to in the anti-spam industry suggested it probably wouldn't), but it probably won't do any harm. The scale of the spam seems so large, that I suspect they'll have to use more automated means of dealing with it.

http://azurelunatic.livejournal.com/6663221.html

Comments

[azurelunatic]

I strongly suspect that there is a bottleneck at available labor to deal with the reports, but I'm working on the following assumptions:
* more data cannot hurt
* some types of spam users go under-reported or not reported at all due to not bothering other users directly (they may get spam email, but delete it without noticing where the links go)
* my suggested method is likely to net some of said under-reported spammers
* it is the ones who are not bothering other users directly who are the primary cause of this specific Spamhaus complaint
* the reports can be dealt with one by one and suspended individually (I suspect the labor bottleneck is here)
* the reports may also be used as a data mine by their analysts

[charmian]

I guess where the person I was talking to differs is that they don't believe that the problem is under-reporting, and that the reports can be dealt with one by one and suspended individually, and they suspected that thing is that LJ can't or won't spend the $ to deal with the problem (through better automation, not through individual labor).

Now, they could be wrong, but personally... I suspect that this really is beyond the scale where they can be dealt with one by one. On the latest posts page, there are 400 posts per minute, and of those, a majority seem to be spam. Let's say 300 of those 400 are spam. 300*60*24=432,000 per day.

[ursamajor]

These kinds of numbers make me adore the Dreamwidth latest page even harder.

Also, when you take into account that userinfo updated_last1 has leapt from ~170k last week to almost 300k today (based on checking LJ Stats a few times on different days: 170k, 180k, 210k, and now 300k; scroll to the very bottom), it makes me think that a bunch of the spammers populated awhile back, and have only started emerging recently. The incredible growth rate just illustrates how quickly the problem is getting worse - because I strongly doubt 40% more LJ users decided to update today-specifically than a week ago.

[charmian]

Well, I think it's hard to compare... If LJ didn't have open registration, they probably wouldn't have these problems.

Yep, a lot of the spamming accounts were created earlier... Pure speculation, but since even before a lot of the accounts on the latest page were spam, maybe what has occurred is that the spammers came up with new tactics to get around the anti-spam measures?

[ursamajor]

Oh, I don't disagree at all. It's apples and oranges at this point, and at least as far as I can tell, having the one week open-registration period for DW hasn't resulted in an uptick in spam. It has made me a fan of invite codes again, though. ;) And also thinking about circles of trust and levels of trust - where they overlap, how far can they be extended, what that means for how populous a given online service can get being invitation-only without being exposed to potential spam-vectors.

And I admit I haven't really looked at LJ's "latest" page in several years; the proportion of signal to noise for me there was already tilted far in the latter direction, probably due to LJ already having a large, frequently-updating population when the latest page was implemented, and I found more interesting things more relevant to me on my friendsfriends page. Dreamwidth's latest page often goes back up to an hour ago; LJ's, on the matter of single-digit minutes. But the ratio's tilted much more favorably to me on DW, and I check DW-latest on a regular basis.

[azurelunatic]

There was an uptick, actually. One account, so far as I can tell.

[ursamajor]

Well, when you start from zero ;)

[charmian]

Anti-spam is basically the only reason I support invite codes at DW. It seems to work, at least.

Hmm, what about circles of trust? To me, it's less about trust than about friction. On DW people give their codes away to strangers, but you can't get a huge number of invite codes at once, and it takes a lot of trouble, relatively speaking, to create an account. This makes it not worth the time of spammers to attack the site (and also, it's small), probably.

I think how populous a site can get simply depends on the demand. I feel like right DW has sort of plateaued, in terms of size, though? Barring some huge LJ mishap.

For years now, LJ's latest page has had a lot of spam on it. Sometimes, depending on the anti-spam measures, it's been better, but then there still is spam.

[foxfirefey]

Oh man, the suspense is better than daytime TV. Is LJ working things out? Or trying to dodge? ONLY TIME WILL TELL.

[azurelunatic]

Huh, I saw that someone in the LJ news comments suggested changing their IP address, as it had worked for them on a situation where they'd been blocked.

My response to that person there was the two situations seemed different -- small outfit, small spam problem, cleaned up, and the organizations blocking him were dragging their feet.

LJ's spam problem is still big, so I fear that if it's being done evasively -- I cannot think of an antispam organization that would be pleased with someone evading their measures without also cleaning up their act. Though I trust that there is cleanup going on. Just, is it enough? I do not want to see LJ get hit with more stringent efforts from antispam organizations.

[charmian]

Hmm, well, if they evade only, the same thing will simply happen again, I guess. And I'm sure they don't want that.

[charmian]

Well, this is interesting. Does this mean the old notifications which failed to be sent will be coming through as well, I wonder?