Shutting Down the Highway to Internet Hell

Do you run a mail server on your home Internet account? If you do, it's probably without your knowledge, such as in a mail worm or a zombie spambot. Few if any people running these programs intend to do so, and it's time for ISPs to close the door through which they operate.

I think there's a consensus developing among anti-spam researchers, many of them responsible for fighting spam on ISP networks, that unrestricted use of TCP port 25 must be shut down to the average Internet consumer. There are those who disagree, but their arguments sound obtuse and defeatist rather than actual justifications to not block port 25.

TCP Port 25 is one of the core interfaces of the Internet, through which Internet mail servers typically send mail to each other. It's normal for users to send data out port 25, but they do so to their own ISP's mail server, from which it is forwarded on to the appropriate location. This is the server identified as the outgoing mail server in the mail client configuration.

But if you are infected with a spam zombie—typically, a mail worm with a backdoor used by a spammer to cause your computer to send out massive amounts of spam—the mail does not go through your mail server. It probably goes directly to the server of the target domain for the spam message. The overwhelming majority of users have no need to do this and are perfectly well-served by sending all their mail through the ISP mail servers. It's also worth reiterating that the block need only be put on consumer client systems, not on higher-end services.

Of course there are users who do need access to the port, or who at least want to run their own mail server and don't intend to abuse the privilege. Or they have a need to use a different mail server than the ISPs, perhaps for reasons involving confidentiality. There are ways for ISPs to accommodate these users.

In fact, there's no reason an ISP can't make exceptions for users who want to use port 25 more openly, especially if they agree to rate limits and to configure it securely. The real problem that needs to be solved is the users who don't know they are running a mail server. Such users won't miss not being able to run one.

Alas, this level of customer service may be too much to expect from some ISPs. Hosting servers are also often far too lax in the management of mail on their networks.

But some ISPs are putting their feet down, attempting to stop the abuse. At the forefront of this effort, defying all conventional wisdom, is AOL. In the 90s, an era of very different circumstances, AOL was the single largest source of spam on the Internet, and the ISP's reputation suffered terribly from it. Now not only AOL users have high-quality spam control, but AOL is perhaps the most active ISP in terms of policing the use and abuse of mail.

Consider the rules at AOL's "Technical Standards for E-mail Delivery." AOL makes extensive use of RBL services like MAPS so that they know to block spam from open relays, spambots, systems with unsecured form-mail scripts and other spam sources. They actually use the same services to block spam that comes directly from residential ISP clients that should not be sending mail directly; in other words, if you don't block port 25 yourself, they will do it for you.

The ISP goes further—much further. If the sending system does not have a PTR record (a reverse DNS), it is rejected. If a message contains a hex-encoded URL (like http://%73%70%61%6d/), it is rejected. If more than 10 percent of the sending system's messages to AOL bounce, AOL may reject mail from it in general. If a server rejects 10 percent or more of the bounce messages sent to it, AOL may reject further connections from the server. There are other, similar rules.

All of this is intended to use AOL's size and clout to make other e-mail administrators set up and administer their systems properly. In many cases, the reverse DNS requirement, for example, the administrator finds out that he or she doesn't have a reverse DNS because AOL blocks the mail, and the end result is an improvement for everyone. Mail servers should have a reverse DNS if they have nothing to hide.

Perhaps not everyone can do everything AOL does. It does, after all, have a proprietary internal mail system. But there's a lot we can learn from its example. Carl Hutzler, until recently in charge of AOL's anti-spam efforts (he has now moved on to a position in engineering and development of AOL's e-mail), has been evangelizing this ethic of responsibility by mail admins, especially at ISPs.

Hutzler warns of the lazy approach of relying on filters, as so many ISPs do. It's the easy way out. But anyone with a little experience knows that filters don't even come close to solving the problem, although they can be a useful part of the solution. I've seen messages with overtly pornographic subject lines and bodies make it through three different Bayesian filters. Spammers know how to play with the content of the message to trick filters.

But the technique that generates the most controversy is when an ISP blocks port 25, as SBC recently began to do.

As one prominent researcher put it, blocking port 25 begins the process of shifting the cost burden for spam from the end user to the ISP and others whose sloppiness in administration is responsible for the unchecked proliferation of spam, and these same people are in a position, through responsible system administration, to choke off most of the abuse. He also argued that the cost benefits of fixing their systems are enough incentive to do it.

Check out eWEEK.com's Security Center for the latest security news, reviews and analysis. And for insights on security coverage around the Web, take a look at eWEEK.com Security Center Editor Larry Seltzer's Weblog.

The depressing counterargument is that many of these systems have excess capacity enough to handle the abuse and that laziness is its own reward. When this is the case, there's no choice but for other ISPs to start blocking the offending ISP, as AOL has done many a time.

This is another point on which a consensus is emerging: that ISPs don't take action to stop spammers on their networks until there is a gun to their heads, generally in the sense that their customers are prevented from sending mail. This is where the major RBLs like Spamhaus and MAPS can play a big role. They have a bad reputation among some, and I've personally been among the collateral damage from an RBL block. But it was my hosting service's fault that my server got on the block because they didn't do anything about the spammer on the same address that I had. Enough of us called and screamed, and something was done about it.

Not every little domain has the clout to block a major ISP. The little guy ends up hurting and angering his customers, but the big ISP won't even notice. But when one major ISP, or a service like MAPS, blocks a major ISP, it gets their attention. The corollary to this is that when you block someone, you need to be responsive when they fix the problem.

The fact that ISPs have no reason to not let users opt out of the system is what cinches it for me. One researcher suggested to me that it was much easier for ISPs just to block a whole range of addresses than to have to put up a system for tracking who was to be blocked and who shouldn't, but this is basically just arguing laziness as an excuse. Besides, the SBC system supports letting users request an opt-out. Why can SBC do it and others can't?

The same researcher was concerned that the opt-out system would be taken over by spammers who would opt-out their zombie systems. But it's not hard to imagine well-designed authentication systems that mail back a message to the customer and require them to connect back.

And as for the added cost to the ISP for this, I'd suggest that they might just save a lot of money by eliminating spammers and mail worms from their networks, but even if you think this is a costly solution, let them charge for the opt-out. Doesn't bother me.

Those who argue against ISPs blocking port 25 generally claim that the downsides are high and that spammers will a) evade the blocks and b) easily move to other techniques for sending spam. Joe St. Sauver has made a well-written case for this position. I admire some of his points, but I still disagree with him, and I think half his problem is that he can't see the point through all his defeatism. Namely, even if spammers were to move to other avenues, it's still worth closing port 25 to stop them from using it.

Getting right to what I feel is the main point, that port 25 blocks will be ineffective because spammers will move to other methods to spread spam, St Sauver brushes aside or ignores counterarguments. He cites recent stories that spammers are beginning to use the ISP mail server instead of sending out spam directly from the client system. There are two counterarguments.

If the ISP requires SMTP AUTH (where you must provide a username and password for the outgoing SMTP mail server as well as the incoming POP3 server), then it will not be a simple matter for the worm to send mail. However, since there are programs available that can read the cached SMTP AUTH credentials from popular mail client programs (click here for one that's sold commercially),

it's not hard to see spam zombies doing the same in the future. They might also do it by monitoring port 25 usage to look for the authentication sequence.

In fact, my own ISP, Speakeasy.net, is very lenient about these things. Speakeasy does not require SMTP AUTH for connections made on their internal network (it does for roaming users), but it says that it monitors mail servers carefully and maintains a number of honeypots on active lookout for malware on its networks.

I spoke to Speakeasy founder and Chairman Michael Apgar, and he insists that a system exhibiting wormlike behavior will not live for long on Speakeasy's network. Within hours the user will be contacted, and if he or she doesn't fix the problem quickly, the plug will be pulled. But Speakeasy is not a conventional ISP; while it's happy to sell to anyone, it has a technically more capable audience who pay more for more open services.

Apgar is quick to agree that mainstream consumer ISPs should be locking down abusable services, and that port 25 is the biggest problem.

Even if the zombie successfully is able to send spam through the ISP mail server, we're still better off than before. The ISP can tell, just by looking at mail server logs, who is spamming from its network. ISPs have a cost interest in fixing the situation and arguably are more responsible for doing so since their own servers were involved. Put simply, forcing the spammer onto the ISP mail server facilitates the elimination of zombies. It also gives the ISP the opportunity to rate-limit mail in general, which will not likely affect regular users, but will seriously cut into spammers' ability to spread the message.

I have a similar reaction to St. Sauver's speculation that zombies, blocked in their ability to send spam, will instead be used for even worse things like denial-of-service attacks. This is not hard to imagine, but while much of the world puts up with systems sending spam, they would feel different about a DOS army. And I can't see that the market for DOS armies scales in the same way that the spam market does. It's just not as big a threat.

He also points out that spammers could still evade blocks on port 25 at the network periphery by spamming inside the network—e.g., to other customers of the same ISP on their subnet. Of course, they will only be able to do so if the recipient mail server is on the same subnet, and this is highly unlikely on a large consumer ISP network.

While most of his writing is laboriously pessimistic, St. Sauver does have interesting constructive criticism. He urges those who would fight spam to focus not on the spam leaving the network but on the traffic coming in to the spambot. He asserts (this is counter to my understanding) that spambots don't typically construct the e-mails they send out programmatically but pass on what they receive from the outside. Whether this is true or not is beside the valid point he makes that it should be possible to look for the command/control coming into the network from spammers. While these commands come in on nonstandard ports, they are known (they have to be, or spammers couldn't find them either).

Finally, for all their claims that easy alternatives exist to port 25, they haven't come up with any. The first port usually listed is TCP 587, but like many of the potential alternatives, it's an authenticated port, so it's not blindly open for spamming use.

In the end, the biggest factor in whether ISPs will play hardball with spammers is whether they want to have to go to the problem of taking out the garbage and keeping their place clean. Some ISPs have complained to me about others who don't seem to care if their networks are used to send out billions of spam messages and mail worms. They don't even look at their own log files!

But the day is coming when these ISPs won't be able to coast through their own laziness and sloppiness. The use of RBLs like MAPS and other blocks of known spammer systems is an increasingly important technique, and if worms really do move to using the ISP mail server, then ISPs who don't do anything about it will find themselves blocked completely by the clean ISPs that are sick and tired of taking abuse.

I don't expect everyone to clean up their act, but think we're moving to an era of unofficial quality standards, of black and white lists, where ISPs will "protect" their customers from the red-light districts of the Internet. It's not perfect, but it's better than what we've got now.

Security Center Editor Larry Seltzer has worked in and written about the computer industry since 1983.

Umm Are you sure about that here is my connection to a mail server

Proto Local Address Foreign Address State
TCP 10.3.8.86:2591 10.2.1.20:25 ESTABLISHED

Yep- that connection shows outbound on 2591 to a host that is listening on 25. The idea of the block is to disallow connections to servers listening on 25.

Ok I read your post wrong, thought you were saying send and receive on 25, this is not that case...

If the spam is being sent by bots closing down 25 wont do anything becuase they dont send on that port..

Actually it's simpler than that- just configure your server to route everything to the ISP's server just as most mail clients do. The problem is that it would not take much time at all for the spammers to get around the block- all they need to do is get the spambot to route all the traffic through the ISP. The SMTP server address can be hacked from the registry. The net result is an exponential increase in the workload on the ISPs' servers with no significant reduction in spam volume.

Fight fire with fire, spam the spammers until their servers blow. Just create a software program that sends the spam back to the spammer 1000 fold for each spam received.

This would quite quickly get your ID terminated... read your AUP (acceptable use policy).

Well, perhaps it wasn't worded too clearly. Any firewall, including the ISP's, can be configured to block traffic that is *calling* any port. So the idea is to block traffic that is calling on port 25 outside the ISP's own network, forcing users to route all messages through the ISP's server.

1) Spam is almost never traceable, it usually comes from a bunk address

Sorry, you're incorrect: mailing headers which contain hostname/IP address are appended to every piece of email at every hop.

The most you can do here is to use a broken system and insert additional, incorrect information before passing along, but the host involved would have been tagged as broken by most RBLs within minutes.

If you run it through the ISP's mail server as client access you may run into mail quota checks, and it's going to interfere with forging the From: address and the headers.

It's not even that complicated. I have SBC DSL and run my own Sendmail server for my own use. When SBC started blocking 25, I simply had to fill out a web form to get it unblocked. Today I still run my own mailserver on 25 as if the blocking didn't even exist...no smarthost required.

Yep, there's no free lunch. To deal with spambots it would make more sense for ISPs to monitor the volume of SMTP traffic and alert the user if there's a spike (perhaps enforced by a block if it continues).

I just found a useful feature in McAfee (I just deployed it here as a replacement for Symantec). You can configure the antivirus client to whitelist the programs that can use port 25, so the only way a trojan can turn your machine into a spambot would be to replace your existing mail client or hack the whitelist. ZoneAlarm has been using that technique for all internet traffic.

If the ISP makes unblocking the port that easy then I'd have no problem with it. My worst nightmare would be having to call the "Your call is important to us" recording to get in re-opened.

The solution requires a number of facets: Introduce an authentication system which requires identities to be globally verifiable...

PKI (public key infrastructure) is intended to encrypt / authenticate email from / to *individuals* not systems and very few places have it running on anything other than a rudimentary basis due to the cost and complexity involved.

The large players could implement Certs for their mail relays, but given the difficulty most locations have with simply running a virus scan on an infected PC, this would also create lots of delivery issues.

And they would be denied until they do.

To my knowledge, the only ISP doing this is AOL and -- to be catty -- a lot of places don't get excited if email for AOL is bounced...

That is, this might as easily be considered a feature :)

Mail is both sent and received via port 25.

Email is received on port 25 -- not sent. You can verify, if you wish, by doing a netstat and checking the ports in use.

Ugh, I agree, especially considering the low quality Indian Tier One support SBC utilizes. Unblocking the port was simple enough, and they had it done three hours after I requested it. It really wasn't an issue.

The only problem I had with it was the way SBC promoted it to users. I didn't know that they'd blocked 25 until I noticed that none of my domains had received any emails for a few days. A quick post to broadbandreports.com revealed the reason, but SBC should have done a better job letting its users know what was about to happen. An EMAIL would have been nice!

Most of the spambots are short-lived on any given machine anyway. All it takes is one recipient that can read headers and he's busted.

Just getting everyone to implement SPF records, and then require a valid SPF resolution before accepting the mail would fix the spambots.

"Most ISPs don't do a reverse lookup but a significant number of corporate mail systems do, which I found out after about 25% of the outbound traffic got hung up in my server."

Yup, you're right. If you don't have that rDNS entry then as time goes by fewer and fewer messsages will be delivered.

Incoming mail is received on port 25. Mail can be sent on any port. 110 and 143 are for POP and IMAP, respectively, which are client applications. IMHO.

You are partially correct. Email is received on port 25, but POP3 and IMAP are used by clients to read email from a hub etc.

Port 25 (SMTP) is generally server to server and POP3/IMAP is client -> server.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.