Decline of spam? You wish!

Via Spamhuntress: Gadi Evron talks about how and why some sorts of spam (by which he means email spam) have almost disappeared. It happened with casino spam and with mortgage spam. And he implies that it will happen with all that viagra spam, too, eventually.

Well, we've seen similar trends in webspam, haven't we? For example, the poker spam that more or less triggered my ventures into webspam fighting has disappeared almost completely (along with The Bulgarians), thank goodness.

So what's left? These figures from my webspam presentation are from March of this year, but are still pretty much correct: The majority of webspam that we see is for pills of all sorts (around 38% back in March) and porn (around 29%). There's a variety of other topics in the one-digit percent values (e.g. finance, OEM software, and ringtones). And then there's a huge portion of about 20% that I threw together as "misc.".

These 20% are what I've identified back then as our future problems: Topics, as Ann-Elisabeth puts it, we never saw before. Spam is branching out. And while the current spam is easy (pills) or relatively easy (porn) to filter, this diversification will make things much harder for us. When spammers start posting about tires or kitchen appliances, it gets increasingly hard to come up with filter rules without blocking legit posts.

Many readers will now point out the existence of services like Akismet that are doing a great job of filtering out even such unexpected spam. However, you shouldn't be too sure that this will continue to work so well in the future. Akismet and friends depend on mass spamming. Only when certain URLs or keywords are coming up with a high volume or frequency will these services be able to reliably flag them as spam (I'm simplyfying things a bit here, but that's more or less how they work). And while the majority of spammers are still operating under the moto "the more the better", some have already gotten the right idea: They are spamming less but over longer periods. Which really makes it hard for anti-spam services to tell them apart from other URLs that may be posted a lot. Keep in mind that information travels fast in the so-called "blogosphere" - a topic and its related URLs and keywords may show up a lot within a short time span when some "hot topic" is sweeping through the blogs. How should a machine know that this is not spam?

And in case you're thinking "CAPTCHA!" now - think again.

Sure, there will eventually be new solutions for all this. What I'm really trying to say here is this:

1. Spam topics may come and go, but spam will be with us for a long time, and in significant amounts.
2. I predict that things will get worse first, and sooner than many people may think now, before they get better.