Top.Mail.Ru
? ?
Previous Entry Share Flag Next Entry
The Bot Wars heat up
device
jducoeur
I just came across an interesting, if not entirely surprising, article on eWeek, which describes how botnets are getting more sophisticated, with more distributed architectures that are harder to track down. So let's think a bit about where this is all going.

(Grounding for those who haven't been following this particular horror: in a nutshell, botnets are where many of the Internet's serious woes come from nowadays. The bad guys who are sending spam, attacking sites, and other such anti-social behaviour, mostly don't do so from their own computers any more, because it's easy for everyone else to track them down and cut them off. Instead, they find computers of ordinary Internet users who aren't keeping their anti-virus software up to date, and subvert them. They install "botnet" software on those computers, so that they can, at the flick of a switch, get all those innocent computers sending out their spam for them. In general, the ordinary users in question don't even realize they're part of a botnet, but they and tens of thousands of others are the ones doing the damage.)

The botnet controllers are getting smarter, with distributed P2P architectures that aren't as susceptible to simply shutting down a few key nodes. The article talks about the fact that the botnet software isn't too hard to find because it uses poor encryption, but I would say that's just a matter of time: just as their architectures have been getting better, the encryption is bound to do so as well. There are limits to how well you can do (when the keys are installed on the subverted machines, it's always possible to break the encryption eventually), but they can get a lot better than they are now.

So here's a question for speculation: what next? One can handwave that the botnet problem would go away if everyone simply kept their anti-virus software up to date, but that seems unrealistic unless it's actually enforced in some fashion. Are we looking at an environment where it eventually becomes illegal to be a *participant* in a botnet? (That is, those unknowing people wind up getting fined for their carelessness.) That seems hard to credit, but I can imagine it happening if a botnet were found to have caused some serious public harm -- you should never underestimate what kinds of laws can get passed in the name of "national security".

OTOH, that doesn't seem likely to actually solve the problem, precisely because it's so hard to figure out who the participating machines are. So perhaps the likelier outcome is the growth of "antigen nets" -- semi-formal networks designed specifically to figure out which machines are participating in the botnets and shut them down. Say you got hundreds of thousands of computers, working together in a distributed network (has to be distributed, or the black-hats will take it down), sharing information about which IP addresses are participating in the problem; this could serve as the basis for a distributed blacklist against those machines. If some major sites became consumers of such a blacklist, it could provide the necessary pressure for people to clean up their acts -- if you find that Amazon locks you out, giving you a message that you need to update your anti-virus before you'll be allowed back in, that provides a rather good incentive.

Of course, this glosses over a *host* of problems with building such a system. The blackhats would surely try to subvert it, planting false positives and trying to remove the true positives, so it would have to be large and powerful enough to statistically resist the bad data. It's not clear whether enough major sites could be made to play along, to provide the necessary incentive for people to fix their systems. Heck, it's probably susceptible to a pile of fascinating legal challenges. But all that said, I don't know what other alternatives there are.

So there's one possible future, maybe 5-8 years out, with the Internet in the middle of a very quiet but serious war between armies of cooperating computers trying to shut each other down. What other possibilities are there? Do you think that the botnet problem is overblown, or is going to gradually eat the entire Internet? Ideas and opinions are solicited...



  • 1
(Deleted comment)
Windows shouldn't activate the network interface intil firewall/anti-virus software is already installed and enabled, or until you expressly click saying that you're off on the internet without a condom.

(Deleted comment)
They've made other security-related changes due to bad publicity. This one isn't a huge deal. And if they really wanted to register within the first few minutes, they could activate the firewall, then the NIC, register, and close them again. If that took as much as a minute it would shave huge percentage points off the likelihood of infection. And they should be able to write a firewall rule that says "connect to MS and send only this amount of info".

(Deleted comment)
If they set it up for momentary activation, they can do it all in an automated fashion. Forcing a user to at least click one more time (acknowledging that they know they are connecting without protection) would at least ease some of the blame and pressure on them. Right now they take the position that they "know best." So prove it. Stopping at stop signs is also an "inconvenience" but most people have been _educated_ to know why they should do so anyway.

(Deleted comment)
...or being shut off by their ISP.

(Deleted comment)
The rise of home networking has actually slowed or stopped much of the spread of compromising worms, just because home routers/NAT devices typically block all unrequested incoming traffic by default now. Almost every OS now ships with a firewall on by default too now. The real risk is after you get set up and start surfing the web. Vulnerabilities in browsers and e-mail programs, and the willingness of Joe Average to install almost anything or to fall for (sometimes cleverly constructed) baited traps is where almost all the invasions come from now.

Vulnerabilities are like Vampires these days, they have to be invited into the system by some willing action on the users part.

For those who have taken the time and money to buy a firewall, that's great - as long as their ISP permits/supports the use of one. Those who don't turn off their firewall because they find it a PITA (note the recent Mac commercials) are also helping.

Many people *don't* have the NAT firewalls, and turn off their firewalls. Some are even using old enough computers that they don't have a firewall installed. And given the willingness of people to open unknown attachments, open and read spam, and install all sorts of strange stuff inadvertently, that's going to be the biggest part of the problem, as it always was. People are the elast secure part of any network.

Permitting firewalls

A lot of people are winding up with firewalls by accident, because they want WiFi. Sometimes the ISP actually supplies the router. My parents got one with their BellSouth DSL, and the CPE for Verizon's FIOS is a WiFi router; you just can't use it without one.

Re: Permitting firewalls

That's *really* good to know. My own experiences have been less good. Setting up my mother's DSL, for example: they told her that she needs to operate without a firewall (or else). When I set up my connection at home, they told me that home firewall use was not permitted; they have since changed. I've heard the same thing from some others; hopefully this is waning.

Re: Permitting firewalls

That is totally weird. I suspect such policies are on their way out, if they're still around at all -- in the current environment, it sounds like pure lawsuit-bait.

It was common to discourage/forbid home *routers* for a number of years, and that's a bit more understandable -- they wanted to sell you three connections, instead of one connection with a router. But even that's mostly gone by now: the fact is, most homes that are even remotely serious are running internal WiFi routers, and even the Comcasts have largely given up fighting that. Besides, it's for their own good: homes with NATs catch far fewer infections, which means they're less likely to be clogging up the network...

The RIAA's legal theory seems to be suspect; there have been a few suits lately where the defendant challenged the notion that an IP address is sufficient to establish liability, and the RIAA dropped the suit rather than set a precedent.

As for blame, there are a number of places to put it (in no particular order):
* Vendors of Operating Systems that are slow to patch known problems, leave open security holes, and penalise people who publicise vulnerabilities.

* Mono-culture architectures, where 95% of desktop machines run one OS (which is known for problems).

* ISPs who refuse to support end users (Cable, DSL) that use hardware firewalls. Some refuse explicitly; lack of a firewall increases your chances of being cracked to a near-certainty within an hour of being connected to a cable-modem.

* Users who don't take care with their home systems - firewalls, anti-virus protection, and of course clicking on attachments and links that do not come from trusted sources.

* Localities and countries without strong laws agains computer-based terrorism encourage crackers, script kiddies and other malefactors to locate within their *virtual) demesne.

Solutions? What is the solution against grass-roots terrorism? Sure, encouraging the fixing of the problems above will help. Education will help. In the end, it *may* avoid the all-out cyber-war that you're predicting. We probably already have some government investment into cracking our enemies' computers; continued investment in how defense may be possible would sure help. I do like the private industry approach you mention, but wonder how it could be implemented without an invasion of the privacy of millions of computer users. Sure, if someone like Amazon does it it may actually have some chance - but what would they do and how?

(Deleted comment)
Financial incentive? That's one of the distinct problems with the idea. If they were vulnerable also, it might work, but if they're careful they don't need to care who connects to them.

Unless people have some sort of Amazon single-sign-on on their machines, which would be ripe for the picking on an infected computer, or unless they turned it into a huge publicity "we're supporting responsible computing" initiative, I don't see how it would work.

Honestly, I don't think it's *that* hard to make a convincing argument. The botnet problem is a huge one for big companies. Hell, it's a big one for small companies, simply due to the risk that some Russian with a grudge could bring down your company. If you could come up with a plan whereby, working together, folks could address this problem, I'd bet that some of the major companies would be entirely willing to join in...

I hope you're correct. The botnet can be death for a company. But spam is a huge problem too, and the big companies do very little that might jeopardise missing an important order, because they want to receive mail from EVERYONE. Some *providers* will use blocking lists to protect their users, but am e-comemrce site can't do that effectively. I'd love to see someone come up with a plan that could not only be effective but be convincing enough; as of yet I'm not that person, but I'm trying... :-)

I do like the private industry approach you mention, but wonder how it could be implemented without an invasion of the privacy of millions of computer users.

I don't think you're quite getting the idea. I'm not suggesting the idea of going and infecting lots of people with this "antigen virus" -- as is mentioned later in the thread, that's already been found to be illegal, and I think it's clearly unethical.

But it ought to be possible to get a large enough critical mass of people *voluntarily* doing this. It requires giving folks an appropriate incentive, but in a world where Bittorrent accounts for most of the traffic on the Net, it's clear that people can voluntarily play together for a good cause.

Basically, the notion is somewhat similar to the one that Blue Frog tried, but fully decentralized so that it can't be brought down the way Blue Frog was -- P2P in structure, and spread all over. A small number of people doing it don't have a chance, but with tens of thousands of machines involved (a number that is entirely plausible), it might be able to collectively start puzzling out the topology of the botnets.

(And yes, the above is full of handwaves. This is a high concept, not a design.)

Sure, if someone like Amazon does it it may actually have some chance - but what would they do and how?

Companies like Amazon might be participants in the antigen net, but more importantly they would be *consumers* of it. The point of the antigen net is to figure out which IP addresses have been tainted. Given that information, it then falls to the sites that people *want* to go to to tell them to clean up their act.

This is, without a doubt, the hardest part of the whole thing: convincing the big players to forego a small amount of short-term revenue (by turning away some customers) in the interest of the long-term benefit of the Net. I'm sure many wouldn't play along. But I'd bet that you could get enough of them to do it to make a big difference. Hell, getting Google alone on board would be huge, and I can see them doing it...

I'm not thinking of active disinfection - that's wrong on a number of levels. My first thought was to (gently) probe the connecting computer to see if it has common virii and terminating the connection if positive, but that's the invasion I was thinking of.

If I understand the Blue Frog concept correctly, you're looking for a distributed database. In this case, you'd list infected computers? I'd be interested in hearing more about such a proposal before I tried to comment; I can see a number of problems with it but don't want to ask about them until I know what you're thinking. I know it's not a design, but I'd like to see the shape your hands are making, regardless.


My first thought was to (gently) probe the connecting computer to see if it has common virii and terminating the connection if positive, but that's the invasion I was thinking of.

Ah -- not really what I was thinking of. I had more in mind paying collective attention to the IP addresses that are doing malicious outbound attacks and probing, and recording them. That's far from a complete solution -- it only works for attacks and probes that require giving away your real IP address (which I believe is true in most but not all cases), and has problems with the impermanent IP addresses of many big ISPs. But it seems like a useful piece of the puzzle, especially if you could get those ISPs involved in the project.

If I understand the Blue Frog concept correctly, you're looking for a distributed database. In this case, you'd list infected computers?

Correct and correct. The problem with Blue Frog was that it wasn't sufficiently distributed -- there was still a central node that could be shut down to take the whole plan out, and the way they were attacked indicates that the spammers *will* get very aggressive if they can find such a central node. But there are architectures (eg, Freenet) that are truly distributed, and extremely difficult to track and attack by design. Indeed, that's exactly what the black-hats are themselves starting to do, for much the same reason.

So the high concept here is having large numbers of machines, working together to build up a distributed DB that creates a statistical picture of which IP addresses are infected. It has to be highly distributed and highly redundant, because the black-hats will definitely try to inject bad data into the mix. And yes, there are *lots* of practical problems with this -- there isn't much beyond high concept here. But in principle, it seems plausible that, if one had many more machines than the bad guys do, one should be able to build up a profile of which ones are causing the problem...

The main problem I see with a large database of infected IPs is (besides dynamic IPs) that it needs to ve VERY responsive. Someone who wants to buy a book at Amazon doesn't want to wait four hours for the refresh, even assuming that they can re-test and re-submit their address as clean in an automated fashion. The larger the distribution of the database, the larger this problem becomes. A DNS change can take hours to propagate even when the change itself is made immediately.

What I find interesting is that a number of large cable ISPs block traffic on specific ports (for example, port 80) ostensibly to block the virus traffic. Realistically, they don't want you to have a server on your machine for fear of actually using your allotted bandwidth, but that's another rant. They don't, though, seem to block much of the *real* problem traffic. Some of this can be dealt with by traffic analysis and management, but they don't want to do that. It may also jeopardise their common carrier status.

Someone who wants to buy a book at Amazon doesn't want to wait four hours for the refresh, even assuming that they can re-test and re-submit their address as clean in an automated fashion.

This is true. Another possibility I've been thinking about is that, instead of blocking the infected addresses, you just *tell* them. My guess is that the vast majority of botnet machines have no idea they are infected.

So one option, with low negative consequences but possibly a reasonably strong upside, is to simply get some major sites giving warnings. If you go to Amazon, Google and eBay, and they all tell you that their records indicate that your machine is virus-infected, that will probably shock a lot of people out of their complacency. Not everyone, but a lot. And if someone remains on the list for a *long* time, they might ratchet up the response gradually.

There are lots of possible options here. I'm under no illusion that any of this is simple or straightforward -- I'm just playing around with this germ of an idea...

Realistically, they don't want you to have a server on your machine for fear of actually using your allotted bandwidth, but that's another rant.

Oh, I'm *well* familiar with this one. Comcast shut down my server ostensibly due to Red Alert, but it's pretty clear that they just don't want me running one. Damned annoying -- I'd far rather run my own system than have to deal with ISPs, at least for my experimental work...

One can handwave that the botnet problem would go away if everyone simply kept their anti-virus software up to date, but that seems unrealistic unless it's actually enforced in some fashion.

Vernor Vinge's latest novel _Rainbows End_ (highly recommended by the by) posits a future where Trusted Computing Platforms are mandated by law, and old-fashioned universal turing machines are considered dangerous contraband by Homeland Security. It's not the main focus of the novel, but it is an important plot element.

Similarly, in Drakon, by S.M. Stirling, computer tech has developed so that there are no universal Turing machines; all programs are read-only, and writing software without a license is a crime. When they make contact with a universe like ours, they're shocked at how easy it is to crack our computers.

During the Sasser/Nimda worm eras, there were actually computer researchers who wanted to release an anti-worm, which would use the same holes the worms were using to get into your system, and then patch them and disappear. The courts ruled that it was illegal to do so, and shut them down. Ironically proving that the laws sometimes deter positive interests while having no deleterious effect on the criminal element. In this case, the ruling was legally right for many reasons, but probably helped the worms overall.

That said, has there been any case where the introduction/evolution of technology has not led to some prognostication of an anarchic wasteland of some sort, be it real, social, or virtual? I don't see the Internet being consumed. I might see the rise of more parallel private networks (a lá Internet2) for safety and/or ensured bandwidth reasons.

Oh, sure. But the reason disaster never strikes is because the systems tend to feed back or die. They survive precisely as well as they come up with coping mechanisms. Those that don't find ways to cope get ground up by the black-hats (think Usenet, which I consider mostly dead by now, due in part to the spam problem). But the Internet as a whole is too important, so *some* kind of coping mechanism must be found.

So think of this posting as precisely that: a suggestion of one possible coping mechanism. Might work, might not, but the pressure to find a solution of some kind is only going to grow...

A Practical Question

How would I -- as a user -- check to see if I were infected? I run a Mac, and I'm generally pretty good about security, so I'm not worried, but...

Re: A Practical Question

I don't know of a sure-fire way, but in general your odds are good if you're keeping up with your anti-virus software. The problem tends to be people who don't pay attention to this stuff.

There are various ways that one discovers that one *is* infected, often having to do with strange processes running, or unexpected network connections happening frequently. But the problem of finding infections is complex, which is why anti-virus software is having to constantly update itself...

  • 1