Tuesday, May 26, 2009

Botnets and Security Hype

A couple of weeks ago a team at the University of California Santa Barbara managed to take over a botnet for ten days. Their fascinating and well-written analysis is well worth reading for an objective and first-hand look at how a botnet really operates.

So how did they do it? A botnet is just the overly sci-fi name for a bunch of computers that are controlled by a central command-and-control structure. The number one challenge for botnet operators is hiding their command-and-control servers to avoid being taken down (the chances of actually being arrested are pretty close to nil). The Torpig botnet uses an increasingly popular technique where client machines try dialling into a set of pre-determined domain names and accept the first server to respond as the botmaster.

This is where the UCSB researchers moved in - they took over the Torpig botnet by sneakily claiming the domain name that was the next in line to be the command-and-control server. The botmasters behind Torpig had not claimed all the domain names that their victims were meant to dial into, either to save money or because they didn't see this coming. In any case, the UCSB found itself in control of a botnet with hundreds of thousands of hosts.

Don't try this at home. The researchers cooperated with law enforcement and other entities to avoid legal problems. This appears to have helped them steer clear of the hot water the BBC found itself in a few weeks ago for actually purchasing a botnet from criminals.

Botnets and the Hype Cycle

You've probably heard botnets talked about on the evening news. Botnets are a particularly successfully marketed part of the FUD-cycle of the information security industry.

But how bad is the botnet problem in reality? Not as bad as previously thought, according to the UCSB team. Previous studies have counted IP addresses rather than actual hosts when estimating the size of a botnet. Getting from IP addresses to actual machines is tough - DHCP leads to an overcounting, NAT to an undercounting, and there are many other factors at play. In the botnet the UCSB team analyzed, they counted 182900 hosts versus 1,247, 642 IP addresses, and there is evidence that IP addresses generally overcount actual machines.

But in many security reports IP addresses and computers are treated synonymously - the latest MacAfee report actually contains the sentence "In this quarter we detected nearly twelve million new IP addresses, computers under the control of spammers and others". Arrghhhh...

Coverage of the UCSB work in the MSM did not mention the overcounting. "Botnets smaller problem than originally thought" doesn't make much of a headline...

So I'm part of a botnet, so what?

Good question. Theoretically, a botmaster could read your email and abuse your other accounts to their heart's desire. In fact, the UCSB researchers performed a keyword analysis of their victims' emails (not sure how they got the legal clearance to do that...). But they are probably the only ones who bothered reading those emails. Botmasters want control of computers to make money and not to read about your date last Saturday. When someone breaks into your house they steal your valuables, not your diary.

Most online accounts and credit cards do not hold their users liable for fraudulent charges. In this way botnets operate a lot like insurance fraud or old-school credit card fraud. They are an annoyance that creates an indirect cost for everyone, but a cost that is sufficiently low that people are willing to bear it. We live in a society where people want to be able to use a 16 digit number they have given out hundreds of times to pay for stuff. If that means that everything costs 1% more to deal with fraud, so be it.

Brian Krebs (who should be on your reading list if he isn't already) posted a piece today about the dangers of allowing your PC to be compromised. Reading through his list of spam, click-through fraud, DoS attacks, and the like, I couldn't get past the feeling of dangerous for society - yes, dangerous for the user - not really. As far as some of the more nefarious password stealing stuff, there is little to no evidence so far that botnets are actually using user credentials for anything other than non-personal misuse of a person's credentials. This isn't great for society, but isn't something the average user is going to care about.

Seems like just the kind of situation that calls for Uncle Sam (or Uncle Barroso)...

Laying Down the Law

The UCSB authors fault registrars for not sufficiently responsive to requests for taking down botnets. While ISP responsibility for content and traffic is a tricky political issue, the content industry has been very successful in forcing ISP accountability for peer-to-peer traffic on their networks. Of course the content industry has a bunch of well paid folks in Washington, Brussels, and other corridors of power pushing their agenda. Botnets do not directly affecting an entire industry's bottom line and so there is no lobbying effort to move responsibility from the client to the registrars and ISPs.

This could change significantly if the national security angle of botnets takes flight. The apparent role of botnets in Internet disruptions during the Russia-Georgia conflict last year, allegations of Chinese cyber-espionage, and frequent stories in the press about the vulnerability of critical infrastructure have attracted the attention of US policy makers. There are even signs that countries like China - long considered a safe haven for hackers - are taking regulatory steps to address botnets.

Regulatory measures will not completely address the botnet issue, but would potentially significantly change the risk/time-invested/reward ratio. Botnets take a high degree of technical expertise to set-up and are of only limited value. A tighter regulatory regime could significantly reduce the incentive for botmasters.

User Education

You often hear about user education in botnet/information security stories, which all too often is vendor-ese for user indoctrination to buy security products. But the UCSB researchers - who have done a great piece of research and aren't selling anything - also focus on user education as a solution to the botnet issue. Their statement that the "malware problem is fundamentally a cultural problem" places the onus for preventing complex and sophisticated criminal activity on the people least capable of preventing it.

It would be nice if all users were capable of being system administrators. For enterprise users, it is fair to expect a minimal level of technical skill. But the truth is that the technical measures a home user needs to take to secure his or her computer are simply beyond the grasp of significant portion of Internet users. The stuff you can educate home users about - choosing better passwords, not recycling passwords, etc is not going to make a real dent in the botnet problem.


  1. How about educating them to not run random programs that they found on the internet?

    Blame the windows software distrubtion model for encouraging exactly this as a 'good idea', or an accepted pratice.

  2. Not running random programs definitely helps prevent some malware. But in the case of Torpig it seems that just visiting an infected site is enough to cause a vulnerable victim's browser to request the attacker's Javascript.

  3. ...which is caused by the sheer difficult of maintaining installed apps on Windows due the lack of a common upgrade/maintainence mechanism.

    Both problems share the same thing in common- the solution.

    Windows needs linux-style package management.

  4. I hear what you're saying, although I think that the situation has improved somewhat at least on the enterprise level with SCCM and other Windows management tools.

  5. .... which complete fails to help home users.

    I don't expect this to happen really, it's probably in the too hard category, with licensing, who you going to trust, would the repo maintainers need to see source, problems.

    That doesn't mean it's not essential to the longer term survival of Windows as a credible desktop platform of course.

  6. Agreed on all counts. I also don't see the Windows maintenance model moving in that direction for the same reasons you mention.

    I am not sure this issue alone can shake the dominance of Windows. For most home users, there are three critical factors in selecting an OS:
    (1) they are familiar with the OS,
    (2) stuff works without configuration, and
    (3) there is a lot of COTS available for it.

    Windows wins (1) and (3) hands down, Mac OS X competes in (2), and Linux - for the nontechnical user - loses all three.

  7. Perhaps- but i beleive what we'll see in the next few years- with OSX too, is the widespread realisation that UAC/User elevation can only take you so far, if you can't trust what you're installing, and you have 100 different update mechanisms you can't stay ontop of it. Malware will continue to be rife- you only need to look at all the ads on google etc, when you type in 'flash' to se everyone trying you to to install something else...

  8. I agree that the update process has become so complex that it is easily manipulated to spread viruses. I just don't see an iPhone-style whitelisting of applications coming to desktops any time soon.


Note: Only a member of this blog may post a comment.