Are Spammers Giving Up? 327
sfjoe writes "Are spammers giving up the game? Google seems to think so. In an article at Wired, Google, '... says that spam attempts, as a percentage of e-mail that's transmitted through its Gmail system, have waned over the last year'. They think their own filters are so good that spammers aren't even trying anymore. 'Other experts disagree with Google, pointing out that overall spam attempts continue to rise. By most estimates, tens of billions of spam messages are sent daily. Yet for most users, the amount of spam arriving in their inboxes has remained relatively flat, thanks to improved filtering.'"
My Experience (Score:4, Interesting)
Spam detection has got to be something like 99.999% accurate
I sometimes get the occasional Nigerian scam letters - but thats it
I've noticed... (Score:4, Interesting)
It seems that have it figured out pretty good to me.
Yahoo (Score:5, Interesting)
Let me tell you, Yahoo!'s spam rate has not improved. I am not sure if their filter isn't as good, or they are just taking money from the wrong people, but I get at least one spam message make it into my inbox per day, maybe 2-3. Oftentimes, the spamming links back to a geocities.com page. Coincidence? I don't know.
With Gmail, I get one spam message per month (maybe) make it into my inbox. They are so rare, its comforting. And since they are so few and far between, I actually use the 'Report Spam' option, because it looks like get this that their filters are actually updated with my input, and I don't see spam of that same type ever again.
This is different from Yahoo, I report spam all the time and yet the same exact message types make it past the filters into my inbox. I even report phishing there, but that doesnt' seem to help.
Can anyone with internal Yahoo webmail operation shed some light into what they actually do with user input? It would be nice to know that someone, somewhere (or at least a script) is using my button clicking for input.
As Much Spam As There Ever Was (Score:2, Interesting)
Don't Filter, Greylist (Score:2, Interesting)
Try http://en.wikipedia.org/wiki/Greylisting [wikipedia.org]
or
http://projects.puremagic.com/greylisting/whitepaper.html [puremagic.com]
Our own office only has about 150 mailboxes but we don't do any filtering at all because of our greylisting as implemented by http://www.openbsd.org/spamd [openbsd.org]
Even better we can greylist at the perimeter instead of letting all of that pointless traffic onto our own network.
And if you're feeling particularily vindictive start posting trapped email address on your own publicly available webpages. Make them invisible or hidden under other content but still harvestable by bots. And soon enough a significant percentage of email addresses out there will point to tarpits. Making botnet spamming a much slower proposition, and should therefore decrease the total ammount of spam.
Re:For Serious? (Score:5, Interesting)
Two Different Truths, But Not In Conflict (Score:3, Interesting)
Well yes, they can easily both be true.
If, for example, spammers are learning that sending spam to @gmail addresses is a pointless exercise in futility. So they further concentrate their efforts on non-gmail addresses.
Google sees a significant drop of spam arriving at gmail (though via accounts which POP3 mail from external addresses, there'll always be some spam).
Everyone else (not Google) sees their inbound spam increasing/strong.
Bandwidth (Score:5, Interesting)
Re:For Serious? (Score:4, Interesting)
Re:For Serious? (Score:5, Interesting)
If the volume of spam grew at X.2% compared to last year's growth of X.9%, that doesn't mean the volume of spam is going down. Hell, one way or another, the volume of spam as a percentage has to go down. It's hard to keep up a healthy growth rate once you've 10 billion a year.
Lies, damn lies and statistics.
Re:My Experience (Score:2, Interesting)
Re:Not giving up, just more Macs and Linux (Score:3, Interesting)
Re:My Experience (Score:5, Interesting)
Re:Spammers give up? Not likely (Score:3, Interesting)
Among the many useful techniques which have been brought to bear against spam from the field of Artificial Intelligence (AI) is the notion of spam as an adversarial game between an intelligent agent (i.e. the filter) and the spammer(s). When this is combined with other AI techniques, such as Bayesian [slashdot.org] or Neural [wikipedia.org] network machine learning type algorithms, the filters become very powerful indeed and not only that but they become automatically adaptable, constantly looking to improve their "score" in the game (i.e. percentage of spams that make it past the filter vs number of false positives) against the spammers. It is important to understand that the creators of this filter do not program the rules but rather the system is designed to perform critical analysis and determine its own rules...this is the power of Artificial Intelligence at work.
Consider that in the past, when serious efforts have been made to bring such intelligent agents up to a high level of play in adversarial games, the programs have advanced to the point where even the very best human players are barely able to win and only with great effort (as in Chess) or, even worse, they cannot win in the face of such tremendously strong play from the AI which never gets tired, never gets psyched out, never panics, but rather constantly and inexorably grinds on to victory with a very high probability.
The spammers are at a distinct disadvantage against such systems for two primary reasons: (1) It is difficult to tell, from the endpoint of the spammer, precisely which message made it through the filter and how and (2) even if they do figure out which messages made it through the filter the filter is learning and training, like the human immune system, for the next time it sees a similar message which will then not make it through. Or in other words the AI filter has full visibility of the game board, but the spammer can only see his pieces and few or none of the pieces of his opponent.
If the game can be made difficult and frustrating enough for the spammer(s) by consistently strong play on the part of the AI filters, then the cost benefit ratio can be reduced asymptotically to zero against the spammer to the point were even the most dogged and determined spammer is tempted to throw in the towel. The cost of sending spam is close to zero but it is not absolutely zero, so the AI should begin discouraging spammers at the point where the AI filter pushes the returns close enough to zero to make spamming unattractive compared to alternative (and potentially more lucrative) activities for the spammer.
Re:I've noticed... (Score:3, Interesting)
Does this sound like a person who will buy your crap? Why try so hard to get around filters in order to reach people who are obviously not going to buy anything from you?
I have NOT seen less (Score:5, Interesting)
2 months ago we received 20 million messages pr week and passed about 800,000 as legitimate mail
Last week we saw 41 million and the same 800,000 passed as legitimate messages.. that's 98% spam!!!
to break it down more..
41 million recieved
32 million rejections on RBL lists
9 million passed onto the spam filters.. 10% of that gets through.
This is for 1 week.
We keep seeing spam double every 2 months.. It's gota stop growing at some point right??
Re:For Serious? (Score:3, Interesting)
Yes, their conclusion depends on some assumptions, but I can believe that those assumptions are at least partly true.
Re:For Serious? (Score:3, Interesting)
Billrocks? What kind of business plan is that? Do they ever pay?
Still not quite right (Score:3, Interesting)
Allow me to correct your correction.
Should be:
"As long as some sucker thinks he might be able to sell something through spam..."
It isn't the general perception of the effectiveness of spam that matters, it's the perception of idiots with dreams of getting rich quick that matter and the supply of said idiots is endless.