Weekly Edition
Return to the Security page
| |||||||||||
Fixing spam with postage
Bill Gates has recently come up with an idea for the spam problem: charge
postage for email. This idea is far from new, of course, but, when Bill
says it, more people listen. On its face, the idea has a certain amount of
appeal. Spammers exist because the economics of the email system favor
them: large amounts of mail can be sent for no money, meaning that even
very small response rates can be profitable. Adding even a small
per-message cost would change the situation considerably. Some variations
of the scheme have email recipients pocketing the postage themselves,
perhaps only if they decide the associated message was unwanted. Others
have ISPs collecting that money; for some strange reason, most ISPs tend to
be more interested in the latter approach.
There are, of course, a few practical problems with this idea. Large mailing lists, for example. If people sending to a list have to pay postage for every recipient, list traffic is likely to drop considerably. If, instead, a message to a list is paid as a single message, large lists will remain attractive targets for at least some spammers. The real problem, however, is that the postage approach, in most implementations, takes a classic end-to-end Internet service and turns it into something centralized. Certainly, one can envision a nice system based on micro-payments where individual users have mail clients which deal with postage issues directly and no central authority is involved. Envisioning MSN or Yahoo choosing such a system is rather harder, however. They will, instead, create a central "post office" which enforces the postage policy and which collects some or all of the money involved. The result is unlikely to resemble the email system we have known for the past couple decades or so. A central post office will require enforcement mechanisms, or people will quickly learn to bypass it. It is hard to imagine unstamped email being easier to stop than, say, music downloads. A postage-for-email scheme looks like a sure way to set off another Internet arms race. Things would be worse if the imposition of a central post office were actually made to work. The temptation to start filtering mail, initially for viruses or some such, would likely prove irresistible. Beyond doubt, the types of mail requiring filtering would grow over time. A central post office would also be an ideal place for governments to apply taxes to electronic mail as their contribution to ending the spam problem. There are also obvious privacy issues to worry about in this scenario. The "postage stamp" approach to spam thus looks problematic on many fronts. Before assuming that these problems would block the acceptance of a central post office, however, one should keep this in mind: the spam problem is getting worse quickly. A great many users will be willing to give up a fair amount of their freedom to somebody who can come up with something that looks like a working solution. This is a scary idea, but it is also a great opportunity. If the free software community can come up with a solution to the bulk of the spam problem while preserving our decentralized net and our freedom, World Domination will be that much closer. (Log in to post comments)
Fixing spam with postage Posted Feb 4, 2004 20:33 UTC (Wed) by csamuel (subscriber, #2624) [Link] Of course, none of this could be motivated by a desire to create a proprietaryformat for all Windows users that would mean that you would have to license the technology from MS to send email would it ? Cynical of Melbourne
Fixing spam with postage Posted Feb 5, 2004 11:17 UTC (Thu) by allesfresser (subscriber, #216) [Link] And of course, when Sir Billy proposes a mail tax, where do you think he intends the cash paid for postage will go? Hmm.... you don't suppose it'll somehow end up near his greedy little digits, do you? Naaaahh...
Fixing spam with postage Posted Feb 4, 2004 21:03 UTC (Wed) by Baylink (subscriber, #755) [Link] IMHO, *the* *number* *one* problem in killing spam is that for any approach to work, it needs to get the buy-in of the "Big 10": AOL, Earthlink, AT&T, Yahoo, Hotmail, MSN, and the rest, in no particular order after the first two. (:-)One such approach that's gaining traction in exactly that space is SPF (does it keep you from getting mailburn?) -- Sender Permitted From. SPF, which checks that the SMTP transaction is coming *from* a machine listed in the sending domain's DNS (as a TXT record), as sort of a "reverse MX", is in fact gaining traction in that important Big 10 space, with AOL among the carriers trying it out. The fact that I found the website for posting here with an unadorned Google search for "spf" probably means something all by itself. My *other* favorite approach to spammers, which I personally think would kill them dead if we could get *it* implemented by a substantial majority of the Big 10, is the variable viscosity tar pit, which Marty Lamb is hidden away in his basement laboratory busily rewriting to make it sufficiently efficient to run in that environment. This one's idea is that the spammier a message gets to look *as the sending SMTP client is sending it to you*... the slower you ack each transaction record. Since the spec for SMTP in RFC 2821, section 4.5.3.2 says that you can delay that ack up to 3 *minutes*... and I can pretty much guarantee you that the bigtime spammers, using software with built-in SMTP senders, have that timeout cranked WAAAAAY back, to like 10 or 20 *seconds*, well, if the message gets spammy enough, they'll give up on you and it didn't cost you anything. *AND*, it may well have tied up one of the spammer's limited number of sending threads. Imagine what things would be like if *lots* of spammers had to wade through that. It'd be kinda like La Brea. "You wouldn't think dinosaurs would get that close to downtown, would'ja?" But hey, maybe it's just me.
Fixing spam with postage Posted Feb 12, 2004 7:31 UTC (Thu) by wnoise (guest, #19404) [Link] Another problem with SPF is that I want to use TXT records for my own purposes.
Hmmm... SPF not the best after all? Posted Feb 15, 2004 9:30 UTC (Sun) by Baylink (subscriber, #755) [Link] That seems to be the argument posed by a couple recent letters to The RISKS Digest -- the most useful website you're probably not following, but should be.Part of their complaint is that the SPF people just went off on their own and did it without following the IETF standards track -- though it seems to me that that isn't entirely congruent with the "rough concensus and running code" standard that's always been in place there: RFC's document *already working* systems, in general. The second writer seems to be complaining that SPF will just make spammers set up real domains which they can SPF themselves... while missing what seems to me to be the perfectly obvious point that then you have a domain you can safely trash all mail from. It's akin, really, to the people who don't understand why forcing edge routers to drop all packets with forged source addresses will help in preventing DDoS attacks, which are inherently difficult to stop: it's because then you'll know where the bastards really are, and you can take preventative measures. And if enough people stomp out packets from RoadRunner IP space, RR's *legit customers* will force RR to take action.
Bigger Problem Posted Feb 4, 2004 22:16 UTC (Wed) by ncm (subscriber, #165) [Link] The bigger problem with this scheme, as with the DNS-based ones, is that, besides its intrusiveness, it wouldn't work. Spammers now hijack ordinary machines authorized to send mail, and send from them. The owners of those machines would pay the postage.
Bigger Problem Posted Feb 4, 2004 22:22 UTC (Wed) by mattdm (subscriber, #18) [Link] The DNS-based ones (SPF, for example) would at least solve a big part of the problem -- replies and bounces would be bourne by the hijacked machine, not a totally innocent bystander who has no way of doing anything at all. It'd be double-bad for the broken-into machines -- but at least that's further incentive for fixing them.Seriously, widely-used SPF plus SpamAssassin *would* solve the massive bulk of the spam problem.
Let MUA and MTA talk Posted Feb 5, 2004 0:41 UTC (Thu) by jarek (guest, #4105) [Link] Today, MUAs (mail user agent) and MTA (mail transfer agent) don't talk much. MTAs receives the mail and stores it in some place where MUAs are expected to find it. If however, the MTA could receive feedback from the MUA about the --quality-- of the email they transfered, they could begin to build a white list and a black list. Remote MTAs not on the white list would suffer delibrately slow response and perpahs even denying certain MTA once their spammer status has been establieshed with certainty to be entered into the black list. With no white- and no black-list, this system would still be operational within the RFCs and boot into fast operation within hours for a large email service and a few weeks for me ;-)At least one good property of this approach is that you don't have to wait for large scale adoption before it works for you. You change it and you benefit. The problem, it seems, would hit large ISPs who would run out of TCP ports. Perhaps they could get started with a good black-list. There is still a problem with myDoom-type viruses if their spread rate is too fast for the black list to be effective, but that may not be a problem in practice. For an ISP which use html-based email clients, the feedback is trivial since they are in control of the MUA. Traditional POP/IMAP mailboxes would require enchancement of that protocol (perhaps, I don't know for sure) and for those of us who run the MTA and MUA on the same machine there's no problem at all. A good thing with distributed MUAs is that many of them run automated spam filtering software like spamassassin (and similar) which is not an option for a MTA due to the resources it requires. This load can easily be accepted by the distributed CPU power of the MUAs and provide fast feedback to the MTA without human intervention for more than 95% of the traffic. /jarek
Let MUA and MTA talk Posted Feb 5, 2004 5:43 UTC (Thu) by freemars (subscriber, #4235) [Link] At least one good property of this approach is that you don't have to wait for large scale adoption before it works for you. You change it and you benefit. The problem, it seems, would hit large ISPs who would run out of TCP ports. Perhaps they could get started with a good black-list. There is still a problem with myDoom-type viruses if their spread rate is too fast for the black list to be effective, but that may not be a problem in practice. This sounds like a Very Good Idea to me, particularly if it includes parts of the variable viscosity tarpit commented on above by "Baylink". Whitelisted email sails through, email from an unknown sender gets slowed a bit (this alone could slow the initial expansion of the next MyDoom), unknown email with spammy characteristics gets slowed by the maximum RFC 2821 allows, and blacklisted senders get rejected. Large ISPs willing to invest some CPU time might take the average of all user whitelists/blacklists into account when deciding how viscous to make the tarpit (but never completely blocking a message the individual recipient hasn't included in his personal blacklist). I like that it's a decentralized solution; there's no centralized authority beyond the ISP to decide what's good and what's bad email. I imaging mailing list software such as MailMan would quickly add some features to cope with clowns who subscribe to a mailing list but fail to whitelist messages from it. The mailer could keep track of which addresses accepted mail slowly and on the next run and would not try to send to them until after the fast ones. Not all mailing lists would benefit, but ones sending biweekly newsletters could get the majority of the emails out quickly, leaving a few days to deal with those MTAs deliberately slowing the connection. One drawback I see is it gives your whitelist to your ISP, creating another collection of info John Ashcroft and friends can demand.
Let MUA and MTA talk Posted Feb 5, 2004 7:44 UTC (Thu) by jarek (guest, #4105) [Link] "One drawback I see is it gives your whitelist to your ISP, creatinganother collection of info John Ashcroft and friends can demand. " Unless ordered by authorities, I don't think the ISP would want to store that information. They would be more interested in one big hash table of good (white) MTAs from which to receive email. If you start to store related information, it will soon become difficult to manage. But the principle of your comment applies. If you give personal information to somebody, that may by used against you (in some sense). /jarek
Let MUA and MTA talk Posted Feb 5, 2004 11:40 UTC (Thu) by dbreakey (guest, #1381) [Link] Check out the latest version of Evolution, which now backends into the SpamAssassin architecture. Not precisely what you were suggesting but certainly a step in the right direction.
... and worms? Posted Feb 5, 2004 2:15 UTC (Thu) by coderock (guest, #12382) [Link] IMHO M$ should first take care of the worm problem, because they're the main reason for it.Only when (hmm... if?) they solve that, they would be allowed to make proposals about other mail problems.
Fixing spam with postage Posted Feb 5, 2004 5:35 UTC (Thu) by zooko (subscriber, #2589) [Link] Adam Back's hashcash (sorry no link -- google it) solves the centralization problem, at the cost of a bunch of CPU cycles.
Fixing spam with postage Posted Feb 5, 2004 5:50 UTC (Thu) by copsewood (subscriber, #199) [Link] Postage carrying mail is as different from SMTP as SMTP is from IRC. I have designed the beginnings of a decentralised and distributed enough architecture that could theoretically do the microtransactions involved, together with a suitably automated currency exchange market. Whether this (MRS) is needed to solve the spam problem is another question. If this approach does gain momentum it would lead to a higher-value type of mail in which you were less likely to lose wanted messaged due to false positive spam detects. It would not be suitable for list server type mail. Certainly people will become a lot less willing to allow their PCs to become spam zombies when every message costs money from their account.
Fixing spam with postage Posted Feb 5, 2004 7:37 UTC (Thu) by lutchann (subscriber, #8872) [Link] I wish people would stop with the "email postage" idea. It is a BAD SOLUTION. Ignoring all the problems mentioned in the article, the premise itself is short-sighted.If we assume we can actually build this postage system thingy, and then make the bigger assumption that it will stop spam without destroying the efficiency and convenience of email entirely, what will spammers do? Concentrate their efforts on instant messaging? Develop better software to create weblog comment spam? Auto-dial our wireless VoIP phones and play recorded messages? Maybe download advertisements directly into our brains with our neural uplinks? The point is that putting all of our effort into designing a system specifically to tackle the RFC822 world would just push spammers into other net-based communications systems, where we'd have to start from the ground up to implement a totally new solution. (What's that you say? We'll just extend the postage system and start charging 5 cents to make a VoIP phone call? What a great idea! And a penny to send an instant message... Oh, no! My Internet bill is higher than my car insurance, because I'm paying for services with totally artificial prices!) What we need is to create some sort of filtering or authentication system that can be applied to *all* Internet communications, and imposes technical limitations rather than economic ones. SPF is a step in the right direction, by adding authenticity to the sender address and accountability for abuse, and the concept can be extended to any IP communications. Once the offenders are identified, it will be much easier to filter them out than when they hide behind anonymous (and unaware) zombie relays. Maybe if the use of public key encryption software was more wide-spread, we could extend the accountability all the way down to the user, giving everybody the capability to filter and blacklist at the level at which the abuse takes place rather than indiscriminately blacklisting entire ISPs. I wish all the hub-bub about email postage was refocused on a solution that would last longer than the keyboard-oriented human-computer interface, and didn't favor economically-advantaged users.
Fixing spam with postage Posted Feb 5, 2004 12:35 UTC (Thu) by wolfrider (guest, #3105) [Link] Ben Franklin put it best: "Those who are willing to give up freedom for a little safety deserve neither freedom nor safety."
Fixing spam with postage Posted Feb 5, 2004 12:43 UTC (Thu) by Baylink (subscriber, #755) [Link] "essential liberty for a little temporary safety". The words everyonemisses are the most important words. I pair it, personally, with a favorite saying that's my own, about the first amendment (and the constitution, generally) "Defending palatable speech is unremarkable." More generally, if it were *easy*, we wouldn't have needed to put it in the Constution.
Fixing spam with postage Posted Feb 6, 2004 0:19 UTC (Fri) by czr (guest, #13701) [Link] I wrote a small document describing the startings of such a system sometime ago, basically it's an authenticated network for communications based on human-level trust. It has both technical and social security measures built in. Since publication, many people have read it but so far none have commented so I guess spam and related problems aren't that important anyway in real life ;-). The document can be found here: http://koltsoff.fi/fbnet I'm experimenting with some code every once in a while but nothing really usable yet. ak.
Fixing spam with postage Posted Feb 5, 2004 10:02 UTC (Thu) by virtex (subscriber, #3019) [Link] I've said it before and I'll say it again -- imposing a postage tax on email will hurt legitimate uses. There are groups out there that send large amounts of email that aren't spam. Consider BugTraq. They have something like 100,000 subscribers. If they had to pay a penny for every email, that would cost them $1000 per message. Multiply that by 30 messages a day and they're up to $30,000 a day just to send email. Also don't forget about the smaller groups. For example, my local LUG mailing list has about 500 subscribers, and is run by one of our members who doesn't make a penny off the service. If he had to start paying $10-$20 a day, he'd have to shut it down.Oh, and before anyone suggests that mailing lists should be exempt from the tax, what's to stop spammers from using a mailing list? Charging a tax on email to stop spammers simply will not work, and I'm disgusted every time I hear someone bring it up as a solution. I'm doubly disgusted to hear someone like Bill Gates suggest it, since too many people will automatically agree with him. I would've hoped someone in his position would actually think these things through, but time and time again, I find that company executives have completely lost touch with reality.
Fixing spam with postage Posted Feb 5, 2004 11:21 UTC (Thu) by copsewood (subscriber, #199) [Link] Yes but how much is a second of your time worth ? You may consider a second it takes you to open a spam and reject it as unwanted as worth nothing or something, but someone else may place a different price on ten of their seconds, if this operation takes them that long, than you place on one of yours. As the sender of a message you may choose to accept or reject the specific micro-price requested for the attention of the person you wish to communicate with.You pays your money and takes your choice.We really are discussing a very different communication protocol here than either fixed centrally-collected postal charges or SMTP mail. No-one will force you to use it - just as no-one forced you to use SMTP mail as an alternative to old-fashioned mail. People will only use micro-payment mail if it offers enough advantages for the early adopters to justify the extra hassles and costs. The same is true for any new communications network. People only used fax because it had advantages for the early adopters, and then network effects took over. If attempts to fix SMTP (such as using SPF) are sufficiently successful, this one is somewhat unlikely ever to be needed.
Fixing spam with postage [not money CPU cycles] Posted Feb 5, 2004 11:46 UTC (Thu) by smoogen (subscriber, #97) [Link] I am wondering if there is some confusion on my part about what the cost is. I think there is a Microsoft research idea that people sending email would pay by 'donating' CPU cycles to the sendee. IE I send a mail message to a relay. The relay asks me some sort of mathematical question. I answer it and then I can send email to that relay. The postage is then collected that way. The more email I send, the bigger the math I do.So it isnt money but the native currency of the NET, cpu that gets costed.
Fixing spam with postage Posted Feb 5, 2004 15:49 UTC (Thu) by vblum (subscriber, #1151) [Link] I would be very happy with Microsoft collecting a fee from its users for every message sent. This feature should be included in their client immediately.Really, if any action is needed: Prosecute the advertising entities. If spam is sent in someone's name without that entity's knowledge, that should be reasonably easy to find out. If, on the other hand, someone does not stop spam-advertising despite repeated infractions, it should be simple to prove that party's guilt by mere specific scrutiny. Authorities do prosecute virus authors and hackers, and take the time to prove their guilt. Why should professional spammers be treated differently instead?
Why not implement a new email system and let the market decide? Posted Feb 6, 2004 7:40 UTC (Fri) by hippy (subscriber, #1488) [Link] I hear all the noise about how terrible the SPAM problem is and I see allthe hot air made about breaking the current email system for the legitimate users by charging a postage fee. What I don't understand is why you need to have either a free-to-send _or_ a pay-to-send system? Why not have both and allow the market to decide. If users really are prepared to pay to get rid of SPAM they can choose to leave the SPAM filled current system and only have an address on the new pay-to-send system. The usual rules of market demand and Metcalfs's law of networks will look after the rest. In the mean time the free-to-send system can work to combat SPAM as best it can and if it succeed the users will come back (unless the pay-to-send system offers enough extra features to make it worth while). I have no problem with a pay-to-send email system, just as long as I do not have to use it. I am prepared to suffer the hassle of running my own anti-spam measures in return for free-to-send email. If others are not, that should be their choice. Richard PS. I get ~150 SPAM a day so I know what bargain I am making.
Fixing spam with postage Posted Feb 9, 2004 4:50 UTC (Mon) by ekj (subscriber, #1524) [Link] Most of the problems mentioned here go away if you replace monetary payment with payment in cpu-cycles."Want this message delievered ?, ok I'll do so if you find some string whose sha1sum ends in 258f." The only known way of doing that is to try, on the average, hashing 32000 strings until you find one that match. It's easy to adjust the postage required. If in some future, spamming is possible with a 16-bit collison (which is what I'm asking for above, "hello\n" would match by the way :-)), you could always say: hash-sum that ends in f2258f Mailing-lists is a problem. But it could be solved by having a white-list of "free postage" servers, and changing the sign-up procedure for a mailing-list to include that the recipient does not demand postage from the mailing-list-server.
Fixing spam with postage Posted Feb 15, 2004 1:07 UTC (Sun) by josh (guest, #17465) [Link] Using a non-trivial calculation is a good idea, but "checksum with these trailing bits" is not a good method, since any such method must be standardized as to how to request the calculation, and once standardized, spammers will be buying "2^20 checksums" CDs for use as lookup tables along with their "1000000 email addresses" CDs. If you scale up the size to where a lookup table doesn't work, non-spammers won't be able to send you an email either. What you need is a problem where any given unit of work is reasonably easy (on the order of a few seconds) to do, there are a huge number of work units so the results cannot be precalculated, and the result of a work unit can be easily verified on the other end.
|
|||||||||||