Researchers unmask two faces of zombie networks
Dr Consumer spam, meet Mr corporate hacking
Botnets are responsible for sending 87.9 per cent of all junk mail messages, according to the latest monthly stats from email security outfit MessageLabs.
A newer botnet, Maazben, first seen in late May, increase junk mail output to account for 1.4 per cent of all spam messages in September, compared to 0.5 per cent of spam in August. Most of the messages promote casino-related spam.
Takedown efforts against rogue hosting firms have removed formerly dominant botnets, such as Cutwail, while other long-standing zombie networks remain a nuisance.
Before the takedown of cybercrime-friendly California-based ISP 3FN in early June, Cutwail was responsible for 45.8 per cent of spam. Grum, responsible for an estimated 23.2 per cent of spam, and Bobax, blamed for 15.7 per cent of junk mail, have filled in the gap left by the effective closure of Cutwail to become the main botnet engines of spam distribution.
Rustock, one of the oldest and largest botnets, accounts for 1.3 to 1.9 million bots and ten per cent of junk mail messages. Spam from the botnet has settled into a regular pattern of starting around 0800 BST, peaking at 1200 BST and shutting off for eight hours at midnight every day. Rustock is the only botnet with a regular spam cycle, according to MessageLabs.
"Over the past year, we have seen a number of ISPs taken offline for hosting botnet activity resulting in a case of sink or swim and an ensuing shift in botnet power," said Paul Wood, MessageLabs intelligence senior analyst at Symantec.
"However, this won’t always be the case as botnet technology has also evolved since the end of 2008 and the most recent ISP closures now have less of an impact on resulting activity as downtime now only lasts a few hours rather than weeks or months as before."
MessageLabs stats on botnet activity reflect what's happening on the internet as a whole. Research by network security firm Damballa, released at last week's Virus Bulletin conference, paint a markedly different picture of enterprise botnet activity.
Damballa found that small (sub-100 member) zombie networks account for 57 per cent of around 600 different botnets it ran into within global enterprise businesses over a three-month period. By contrast, huge botnets (10,000 strong or more) accounted for just five per cent of the malware-infected PCs networks spotted by Damballa.
Not only are smaller botnets much more commonplace in the workplace but they are geared towards different malicious actions. Most small business botnets are established using popular DIY malware construction kits, such as Zeus and Poison Ivy, and are being used for targeted reconnaissance of corporate networks or in order to set up backdoors on critical systems.
The targeted hacking application of business botnets contrasts with the use of massive botnets to distribute spam or launch denial of service attacks.
More on Damballa's research, along with a pie-chart on corporate botnet size, can be found in a blog posting here. ®
@ Charles 9 and Ken Hagan
Yes, traffic can be sent FROM any port, but must be sent TO port 25. Some ISPs (to their credit) block outbound connections TO port 25 so that their customers cannot send mail - unfortunately some of them don't have the ability to lift that block for customers that do legitimately send mail directly. Normally the ISP will provide an outbound relay for it's customers to use, so it isn't a big problem for most customers - and the outbound relay can impose checks to detect unusual traffic patterns.
Many mail servers intended for relaying of mail will accept connections to the "submission" port (587). Since these connections *should* require authentication, then it's no good spammers trying to use them, but they allow users to send even when behind a firewall that blocks access to port 25.
Doesn't SMTP only need to send TO port 25, not necessarily FROM it? Blocking a user's port 25 is useless, then, as the bots would send their traffic FROM other ports. As for blocking all traffic with a port 25 destination (other than the ISP's designated SMTP server), that won't work either since rogue SMTP servers don't have to abide by the standard of using port 25 as the target. ANY server can use ANY port since there is no enforceable means to block otherwise--nothing there but rules.
I may be about to have a senior moment here, but presumably these machines that are sending mail are sending *to* port 25 on a machine that *has* a proper MX record (since spam is sent to conventional email addresses like firstname.lastname@example.org). Apart from the contents, this is surely indistinguishable from the process involved in sending legitimate email from the compromised machine. Similarly, botnet control can be done over http, with the compromised machine polling a range of controllers rather than listening for instructions.
I can't see why a botnet member would need to be listening on any port. Blocking outgoing traffic *to* ports 25 and 80 will upset the average punter (who only uses the machine for email and web). OTOH, deep packet inspection of the aforementioned traffic will leave all the non-average punters (who read El Reg) absolutely incandescent.
So what are the ISPs supposed to do?