50 ISPs Harbor Half of All Infected Machines

Orome1 writes "As the classic method of combating botnets by taking down command and control centers has proven pretty much ineffective in the long run, there has been lots of talk lately about new stratagems that could bring about the desired result. A group of researchers from the Delft University of Technology and Michigan State University have recently released an analysis of the role that ISPs could play in botnet mitigation — an analysis that led to interesting conclusions. The often believed assumption that the presence of a high speed broadband connection is linked to the widespread presence of botnet infection in a country has been proven false."

Comments

Duh.

[TaoPhoenix]

Well, since Verizon and Comcast harbor 10% of all user customer PC's all by themselves, this is not so impressive.

Re:Duh.

[realityimpaired]

I'm guessing far fewer than 50%... while I could be wrong, the point they're trying to make is that a handful of small ISP's which don't seem to pay attention to security are a major source of the problem.

While I know it'll have a bunch of the net neutrality folks up in arms, it's relatively trivial for an ISP to redirect all outgoing traffic on port 25 through their internal mail servers, and to run server-side anti-virus on all outgoing mail. They can go one further, and rather than blacklisting potential viruses, they can work off a whitelist of allowed senders (sender e-mail address, in the case of my ISP), and require secure authentication to relay. My own ISP does exactly that, and while somewhat draconian it doesn't really affect the average user, and, when coupled with a blacklist of known viruses, it does take a significant chunk out of the potential to cause harm to others if you get infected yourself.

Re:Duh.

[mikael_j]

Unfortunately I've worked for several ISPs that had the bad habit of enforcing the following:

• Blocked outgoing connections on port 25 for all hosts except their own SMTP relay.
• Required valid logins on the SMTP relay in order to send emails.
• Draconian size limits on emails passing through the SMTP relay.
• Low upper limit on number of emails per day through the relay.
• Antivirus software that ripped all sorts of benign data from emails for no reason.

Let's just say there were plenty of issues with users who couldn't figure out how to set things up on their own, not to mention users who found out the hard way that large attachments caused their emails to bounce (somewhere in the 10-15 MiB range IIRC).

Personally I'd love if there was at least an option for completely unfiltered access (perhaps even proper reverse lookup to deal with the idiots who think reverse lookup is a good way to deal with spam (hint: it's not, way too many legit companies have multiple hostnames on their mail servers or use a third party's mail relay for this to work well, it just gimps email)). Now. I'm not saying this should be for everyone, filter by default but give users an option to turn the filter off completely but display an overly clear "don't do this unless you're absolutely certain you know what you're doing" message that includes a warning about how the ISP will shut them down in a nanosecond if they get any legit spam reports. That way those who really want/need unfiltered access can have it while the rest of the users can enjoy the walled garden.

Re:Duh.

[mikael_j]

Why would you want to send mail from a residential IP?

Because it should be possible.

The vast majority of big mail servers will simply block your messages.

I've found it's more like a minority, and I've even encountered a few that block large swaths of IPs that they have tagged as "residential/dynamic" but will let incoming emails through if there's a proper matching SPF record.

What's the point of email if you don't have reliable delivery?

It's only unreliable because some admins are lazy. And boy, it sure is fun when an IP that's been a static business IP for years suddenly gets blacklisted as "dynamic residential"...

If you want to access your own mail server running elsewhere, it should be trivial for it to allow inbound connections requiring smtp auth on a port other than 25.

It's still just a workaround that doesn't need to be done if the ISP handles its network properly instead of just randomly blocking ports for shits and giggles. And most only block outgoing port 25 so it's pretty easy to set up your MTA to send via their relay and run the MTA locally anyway, but this still retains the problem of the ISP filtering and messing with outgoing email (as well as the potential loss of outside access if their SMTP relay decides to go down, and I've seen enough ancient Solaris machines handling customer email to have a strong distrust of ISP SMTP relays, it shouldn't be "normal" for it to go down at least 1-2 times per week if you have tens of thousands of customers).

Re:Duh.

[KingMotley]

Spoken like a gmail/yahoo/hotmail web user. Sorry, I actually use a real email client, and send/receive emails to and from multiple email accounts all from my one email client.

See there is this thing called an email standard, and that standard specifies port 25 is used for that purpose. Maybe a better standard needs to be made, but until then I want my ISP to leave port 25 alone. If they catch me sending spam from it, feel free to send me an letter and email and block the port temporarily.

Obvious solution

[qbast]

"I say we take off and nuke the entire site from orbit. It's the only way to be sure."

Wrong way of looking at the problem

[Rosco P. Coltrane]

The real shocking truth here is that one single OS harbors the vast majority of botnets and viruses. That OS should be the real target, not ISPs or poor users or something. Sheesh...

Botnet sans broadband? Seen it already...

[damn_registrars]

My site at home has been under a distributed hack attempt (a long list of IPs all trying to ssh in as root*) for days now. On the first day the attempts were quite frequent; approaching 1 per minute. Now on day 4 the attempts are trickling it as infrequently as one every 20 minutes. A system on a reasonably fast connection could on its own surpass the 1/minute barrier when running a dictionary password attempt through ssh if it wanted to; hence this looks like it could well be systems on slow connections. Add in that some IPs disappear for a while and then come back - as if the PC is logging off and then on again - and it certainly does look like a low-speed botnet.

* Naturally, my ssh denies all root attempts. Even if they got the password right they wouldn't know it, because the rejection would be the same. Other botnets have tried whitepages-style attacks using long lists of common user names and not matched any allowed users on my system as well.

** Yes I know I could just change my ssh port and much of this would go away. But I find it amusing and I have bandwidth to burn.

very flawed logic

[frovingslosh]

One big problem with this logic is that it is based on IP addresses analyzed from captured spam. The problem with that is some major ISPs (including AT&T) are blocking access to out-of-network e-mail servers, and doing other things to make it difficult for even their legitimate customers to send legitimate e-mail. So this method of knowing where the botnets are would completely miss major botnets if they are unable to get spam out efficiently.

You may say "Why does that matter as long as the spam is stopped?", but it matters a lot. The machines are still infected and could be used for other things, from denial of service attacks to hosting and spreading kiddy porn to just watching for private data to go by (like banking information and credit card numbers) and report them directly back to the control system. Making major judgments about botnets based only on IP addresses seen in spam is short sighted and foolish. And it also assumes that all botnets are honest enough to not forge IP addresses. Any smart botnet could easily forge the IP address the spam is coming from, to make it that much harder to find. If a clever bot just changed the fourth or even third and fourth part of the IP address and replaced it with a random number, the botnet would look much larger than it really is and make it much harder to track back to the infected machine, but would not be easy to detect by comparing the supposed source IP and the SMTP server from outside the network.

Who are they?

[HangingChad]

"The networks of just 50 ISPs account for around half of all infected machines worldwide," say the researchers.

Who are the 50? Publish the names and IP ranges and let the admins loose on them.