Hunting for Botnet Command and Controls

Uky writes "Convinced that the recent upswing in virus and Trojan attacks is directly linked to the creation of botnets for nefarious purposes, a group of high-profile security researchers is fighting back, vigilante-style. The objective of the group, which operates on closed, invite-only mailing lists, is to pinpoint and ultimately disable the C&C (command-and-control) infrastructure that sends instructions to millions of zombie drone machines hijacked by malicious hackers." From the article: "Using data from IP flows passing through routers and reverse-engineering tools to peek under the hood of new Trojans, Thompson said the researchers are able to figure out how the botnet owner sends instructions to the compromised machines."

Botnet

Now only if they could do this with Skynet, we might just be able to postpone Judgement Day another 6 years.

Uh oh!

uh, time for me to disconnect to the internets for a while.

Easy way to catch them.

Easiest way is to create a small IRC network, and submit the name to all the irc clients out there, so it'll be in the list. Also, name it something so it appears at the top or near the top...

To inflate user counts, just get an ircd that allows assigning yourself or others fake hostnames (for certain hosts/etc). Then load tons of bots in channels pretending to be 'users'. You could even get creative and make them idely chatter with each other..

Anyways, the point is that most of these botnet peoples eventually want to take a part of their net out to go mess with irc channels, and they usually seem to target smaller networks on the top of whatever list they're using.. So all ya gotta do if just log massive joins into certain channels, or when a flood of users magically connect to your fake network.. Then you have tons of bots to dissect or whatever.

C&C attacks work well for military

C&C attacks are the staple of today's military. An organized, centralized effort should do wonders for laying waste to the economic value (and motivation) behind such behavior.

Re:C&C attacks work well for military

C&C attacks are the staple of today's military. An organized, centralized effort should do wonders for laying waste to the economic value (and motivation) behind such behavior.

The best way to lay waste to someone's economic power in C&C is to destroy their harvesters. Make sure not to send infantry units because they'll suffer tiberium poisoning, or merely be run over by the harvester. Another great way to wreak havoc is to send the engineer into the harvesting facility as the harvester is unloading, you'll get the building, harvester and the tiberium thats being unloaded at the time. Of course, many believe engineering cheese is the cheap way to play C&C, but of course there are too many cheesy plays to count in that game. I suggest you play something like Starcraft. Or Starcraft2, which I have a chance of actually helping with.

Violation of My Privacy?

"Using data from IP flows passing through routers and reverse-engineering tools to peek under the hood of new Trojans, Thompson said the researchers are able to figure out how the botnet owner sends instructions to the compromised machines."

When the security "experts" are busy looking at all the data passing through routers, who is busy ensuring that the "experts" will not violate my privacy by reading the personal but sensitive e-mail notes that I send to my friends and associates?

In other words, when the "experts" are protecting me from the hackers, who is protecting me from the "experts"?

Re:Violation of My Privacy?

When the security "experts" are busy looking at all the data passing through routers, who is busy ensuring that the "experts" will not violate my privacy by reading the personal but sensitive e-mail notes that I send to my friends and associates?

You, by encrypting them.

Re:Violation of My Privacy?

At every company/ISP there are people who have the ability, and regularly do, delve into the data streams flowing through the routers. And yes, sometimes they read your letter to Aunt Martha (or worse).

Mostly the volume of data involved is so large that trying to monitor it without filtering for the items of interest is usually impossible. And that filter is your best defense, in this particular situation.

Unless, of course, you're sending Aunt Martha that e-mail over IRC....

Re:Violation of My Privacy?

Does it come as a surprise to you that people that have access to routers can sniff your packets?

Re:Violation of My Privacy?

When the security "experts" are busy looking at all the data passing through routers, who is busy ensuring that the "experts" will not violate my privacy by reading the personal but sensitive e-mail notes that I send to my friends and associates?

Umm.. they're not looking at "all the data passing through routers". Flow data is a sampling of information (source, dest, proto, port, etc) from a designated collection point. Even without the actual "data" portion of the packet, it's impractical to collect anything more than a small percentage of the total traffic.

So you can put the gun down- your privacy is safe.

Re:Violation of My Privacy?

I've owned a couple of ISPs and I currently do service for a regional provider. If I cared to look I could see everything - your best defense is the same reason that you don't get dates - what you do is just not that interesting to anyone else.

pessimistic

So is this news something to be pessimistic about or what? As I understand it, without vigilantes botnets would be even more "unstoppable" than they are now. It's cool that they're mitigating it, but it really comes down to getting some cooperation going on multiple levels... starting with the ISPs acting more against outgoing malicious traffic for a start.

Shutting down botnets is a pointless effort..

The problem isn't botnets, the problem is people and systems. The only reason botnets exist is due to the fact that current software is engineered without much thought toward security, and vendor supplied patches are not applied. Shutting down a botnet is at most only minimally worth the effort as the hosts are still vulnerable to be aquired by the next virus that comes around.

The only solution is secure software engineering and prompt, reliable patching.

Re:Shutting down botnets is a pointless effort..

and until then we'll just let the botnets run rampant....

Unfortunately that's not a very good solution. While creating more secure software from the ground up is definately thew ay to go for the future you have to have some plan to deal with the current problems. Keep in mind that the vast majority of people aren't going to upgrade to the latest and greatest OS, web browser, or whatever if thier existing one works. So even after you've got more secure computing solutions out there you have to convince people it's worth the time and more specifically, cost, of upgrading.

kudos

it's great that industry, when faced with a lack of effort from the law and legislature, has the will and wherewithal to go after the scumbags. it's a great first step to show policymakers how much of a concern this is to internet security.

Told Ya So

Internet ages ago, when DDOS was hot and researchers all concentrated on that threat, I tried to tell them that DDOS is nothing. Stuart and the others wrote their paper and based the threat on DDOS which influences computer security research even today. I predicted what is now called botnets would be the more frightening destination of the DDOS train. I didn't catch that IRC would be the covert channel of choice (not very covert). HTTPS seemed much more likely to me - net admins expect to see https traffic.
The vigilantes are running into the problem of cut-outs. The original botnets for DDOS all used a three-tier architecture - slaves (bots), masters (IRC servers), and clients. The current incarnation seems to have at least that many layers if not more. Killing the masters is better than trying to stomp on all the bots, but that still leaves the clients. Until the owners of the compromised boxen acting as masters allow access to track back to the clients, the vigilantes are facing the fate of Sisyphus.

Goetz - AC because I can't remember my /. user name

Self destruct the botnets?

What would be really interesting is if using a combination of honeypot PCs (to match trojans to controllers) and the commands used to control the botnets, these vigilantes could make the zombified PCs download and run a cleaning tool to rid themselves of the trojan.

What causes botnets?

Well, obviously script kiddies with the malice and idiocy to create them. But also, the end users ... the people who irresponsibly leave their machine open to the 'net, get 0wned, and then contribute to whatever DoS is going on.
These end users just *don't care*. Because, although their owned boxen are f-ing with the rest of the internet, it doesn't affect them - a selfish luser attitude, why should they bother virus/trojan scanning their boxen?
I wish ISPs would hold the lusers (criminally) responsible for this. I for one look after my home datacentre, including my Gentoo Linux boxen and keep them patched.

Good for them.

From the FAS:

a group of high-profile security researchers is fighting back, vigilante-style.

This emotionally laden language has been deliberately chosen to make it sound like this activty is a "bad thing [tm]"

I truly believe it is the duty of every person to fight against clearly evil activity.

This includes a mugger hitting an old lady, a middle age man trying to drag a pre-teen girl (or boy) in to a car idiling in the street, and a person trying to kick in the door of the elderly couple down the street.

If the people disabling bot-nets make every effort to be certain they do not harm innocent or uninvolved people (and the standard here is very high), then they are doing a public service. (if they take the attitude, like some "anti-spam" people, of -> 'kill them all, let God sort them out, they are just assholes with very, very small peckers')

Those who believe the gub'mint is going to be johnny on the spot to fix all your boo-boos are sadly misguided: there is neither the manpower or the reaction time to fix everything "bad" in the world. That depends on YOU.

A more effective approach?

I'm wondering why they aren't telling bots to self-destruct? It seems pretty obvious to me that the C&C structures could reform fluidly as you take them down? A Black hat has a list of his bots, if you nuke his IRC channel, he just spawns a new one, or moves to a new IRC network...

But if instead you tell all his bots to wipe themselves out, he's got to buy new ones. Yes those machines will surely get reinfected within a few days/weeks, but it will throw a much bigger wrench in the works.

How is this not the obvious approach? Why aren't they doing it? Or maybe they are and aren't stupid enough to tell the media....

Vigilantism? Or good citizenship?

The word "vigilante" keeps getting bandied about by people who frankly probably just sit on their backsides all day and do nothing to try and help the problem.

The botnets represent a serious threat in all sorts of different ways. Spamming. Phishing. DDOS attacks. Extortion. Money laundering. Child pornography. These large armies of zombie PCs can be use for a variety of evil purposes.

Yah.. this should be the remit of law enforcement agencies.. but guess what. Nothing much is happening. Law enforcement is either waaay outta their league or swamped with other issues. So as good citizens of the internet, what should we do?

Well.. those people who keep moaning about "vigilantes" will do nothing.. expect moan some more when their business is taken out by a DDOS-wielding extortionist. One basic obligation of all citizens it to protect others and to not ignore crimes when they are in progress. So, it is absolutely right and proper that people take direct action if it is clear that law enforcement agencies cannot.

You can target the botnet's C&C system. And there are a variety of ways you can do this - not all of which require immense technical skills. Sometimes that means you have to be slightly more "grey hat" than "white hat" in your approach.

But even if you are technically breaking the law to shut down a botnet.. exactly *who* are the victims? Nobody important, that's who - and they are usually hiding behind layer upon layer of false domain registrations, hijacked IP addresses and worse. In fact, most of the time there are no identifiable victims of this type of anti-botnet action at all - no valid names, companies or organisations. So who's gonna complain?

Personally, I'm not part of this group, but independently I have managed to shut down two large botnets.. at least temporarily. And I would do it again. But.. well, let's just say if you are involved in this sort of thing then it's better to stay an "Anonymous Coward".

I hope they invite the DShield guy

I can't find it on his site, but the guy who runs DShield was under a DDOS attack a few years ago and he managed to crack into the IRC channel the attacker used to control his bot network.

Apparently the attacker about crapped his drawers when instead of the usual bot replies to his commands an actual person started talking to him in his IRC channel.

http://dshield.org/ [dshield.org]

The new superheroes...(whats their name?)

... fighting back the internet scumbags all over the planet, vigilante style...

Now if they could just have a cool name, we could have a new hit superheroes movie for this summer.

Any suggestion anyone ?
- The League of Net Shadows
- The League of Extraordinay Nerds
- The Fantastic Fourty

Come on give me something better ...

Re:The new superheroes...(whats their name?)

In honor of one of the common infection vectors: The Active X-Men.

Of course, the need to acknowledge both genders would probably make Active X-Force or Active X-Factor a better choice.

What's good for the goose...

So, how is this different from a "Star Chamber"?

I'd be interested to see how many people in /. who might applaud this pro-active white-hattery, who simultaneously strenuously object to the US Patriot act which is pretty much just allowing the government to do the same thing in real life?

book deal

THey should write a book about this. It will be like a modern day Cuckoo's Egg.

Since malware writers aren't held liable...

...and likely because their wares are useless until activated by an idiot enduser but mostly because government neither is competent enough to go after this or should be trusted enough, then I don't see why extending antibodies to the malware problem doesn't deserve a shot.

With honeypots and careful use of infectable machines, the code that makes up these beasts can be examined and anti-malware can be released into the wild to destroy the infections whereever the anti-malware gets installed by an end-user.

"Wow, I just cleaned spyware off my machine by looking for pr0n." Sort of like accidentally giving yourself life-saving medication because someone knew you were a pill popping idiot and they put the right stuff where you'd find it.

The question is, how would the corporate antimalware forces of right now react? "Symantec finds the W32.SpamZapFly2 to be a highly dangerous worm capable of closing far too many open smtp relays (which is eating into our business) and recommends using our new tool to remove it as well as purchasing our latest antivirus software (which will be as ineffective as the last one) instead of relying on accidental infection with this so-called anti-bodyware (because while it has equal chance of happening, we'd prefer to be paid).

C&C?

I thought there was no such thing as a central C&C on botnets. An infected pc, can be a member of many botnets. Today a pc is doing the bidding of joe hax0r, tomorrow is doing the bidding of billy rox0r. Even if you shut down one C&C, the thousands of infected pcs, remain infected and ready to join another botnet.

The only sollution is user education.

Re:C&C?

I thought there was no such thing as a central C&C on botnets. An infected pc, can be a member of many botnets.

Yes, but there'll be one trojan per botnet. Script kiddies don't like to share, and in fact the current trend is supposedly groups assembling botnets and then auctioning off their services to spammers. Given that, you can see why the botnet "owner" wouldn't want to allow access to other evildoers.

Typical freeloaders

Using data from IP flows passing through routers and reverse-engineering tools to peek under the hood of new Trojans, Thompson said the researchers are able to figure out how the botnet owner sends instructions to the compromised machines.

This is a blatant violation of the trojans' EULAs if I ever saw one. The authors put a lot of work into writing those trojans. What gives "security researchers" such a sense of entitlement to that code? If they want to analyze malware, they should write their own!

westwood studios

seeing C&C in the article took me back to the good ol' days of DOS and Westwood Studios...

fuckin Electronic Arts.

Anti-anti-botnet

Once the head goes, that botnet is largely useless," said Roger Thompson, director of malicious content research at Computer Associates International Inc.

If I were a Blackhat, my counter to this would be to have the members of the botnet relay my commands among themselves like a telephone relay tree where one person calls 5 who each then call 5 who each then ... To find Mr. Big, you'd have to find the headwaters of the stream, which would be a difficult task.

Re:Anti-anti-botnet

No point in treeing it, trees lead to an origin too easily. Cell-style works so much better. Each peer has to discover eachother (Start with the machine that infected it, get the current list of peers from it. randomly ping each peer to see if one drops off, if so send a hint to your other peers. All hints only cause verification, not actually removing. Same for adding new peers this way.
Controlling it is then a matter of keysigned commands. All commands are timestamped to be unique(so you can easily discard duplicate messages), and is verified with the public key. The only way you can be exposed at the leader is if you get caught with the private key.

..just another wasted effort.

I'm sure the research they're doing is quite intriguing, but when are we going to stop wasting time and money on clueless users that obviously don't understand the importance of clicking on "Windows Update" every couple weeks? That type of negligence is unacceptable in almost every other facet of everyday life, so why aren't businesses holding the actual users responsible for the damage they cause?

You know what I'd like to see for once? How about a vigilante team of lawyers that work with the businesses who regularly pay $many thousands of dollars to deal with these botnet attacks. Work with the ISPs involved to get customer names, through *legal channels* (so they'll cooperate), and then file civil lawsuits against every single one of negligent computer owners. Hmm, what's that? Oh, now you've got everyone's attention.

Even a DirecTV/RIAA-style letter campaign would probably do more to fight the problem than this team of bot-busters will.

Note: no disrespect to the research team. I applaud their efforts, even though I think there's a better solution to this problem.

No sniffing needed

I decided to go botnet-breaking a few months ago (when someone "invited" my webserver)-- no need to sniff stuff. Simply edit/clean the botnet script and run it. Take note of the hostnames used by all the other bots in the channel and begin scanning them for contact info. Usually SMTP or such will give a hostname and from there you just contact the admin of it.

New spin on...

Sounds like a new spin on something Steve Gibson did a few years [grc.com] ago. Very interesting read.

Not a problem for long...

Hey guys. Just thought that I would put my $0.02 in.

I am not into botnets anymore, but like most here prolly', I started my internet life on irc. And anyone else who grew up on non dalnet like servers with chan services knows that being on a network without them can be a pain. Especially when smacktards show up for the day ;)

Anyways, knowing a bit about bot's and botnets, I would say that it shouldnt be too hard to take some down. Being irc based, plain text would be one problem. But if you have access to a machine infected, encryption would be pointless since you could just debug the program and find out what it 's protocol is anyways. I think one big issue that was hinted at in one of the above posts was that you should be able to use an infected machine to "take over" the botnet. Well, things dont work that way. For those of you that havent run one or used one before, I will give you a rough idea of what the ones in my day (1.1.15 or so IIRC).

A botnet is basically a shell like environment similar to say a bash shell or a dos prompt. ie: its all text commands using plain ol' ascii. Commands generally start with a ".", like ".help". The botnet also has security systems in place (ie: users with passwords etc) that define who can dcc chat the bot directly, use its !channel commands on irc etc. The eggdrop (sorry, yes, im refering to eggdrop's specifically) bot also has the ability to link multiple bots togethere to form a big "botnet". The is all of course done with special bot accounts with unique passwords.

The reason you cant just take one over (despite it probably being a modified version of this system of bot), is because the other bots are probably only allowed to "take orders" from a specific machine or user. Although for simiplicity sake, I would imagine its just a user and password combo to prevent any traceable information from being gleamed over the botnet traffic. Dont forget to that the botnet would be point to point and most of the traffic would only be coming from a single location (which you would have to find out from a comprimised machine).

In the end, I see the biggest problem in finding the zombies being, how do you tell when a machines infected if the virus tries the best it can to hide itself from non-forensic integrity checking tools. But, over the years I can see software taking a turn to being better checked for authenticity and integrity etc. Once we hit that point, botnets would probably start to disappear. Also consider that the machines themselevs will go offline and be replaced by newer ones that arent suceptable to the same malicious code. This at least forces them to keep active. And keeping them active helps you trace them.

Anyways, hope you had a fun read. Not worth previewing this one, l8r.

How my botnet would work.

If I were a blackhat, my botnet would run thusly:

The bots would be connected to their own P2P-ish system. Commands would be passed around the network in a method similar to searches in Gnutella.

All commands would by signed by my private key. My bots would all have my public key. This, I would be *the only person* who could issue valid commands to my botnet.

This would make it impossible to tell where the commands are coming from since the originator would look just like another bot on the network.

Why allow IRC?

I may be missing something here, but if IRC is used to control malicious programs, then why allow IRC?
Call me a stick in the mud, but I have simply never seen the purpose of IRC. I've installed programs for it, logged into the LUG's channels because I'm told it's the best thing since sliced bread, found it to be a an utter waste of time, and removed the IRC client. Three times. I simply can't see any purpose to it that is worth either the massive time waste (people don't think before they reply to questions), or the huge security hole that it appears to be. [BTW for people on AberLUG, I know there's a no-install Java access route too. But there's no content.]
So why are people (network administrators, specifically) allowing the packets to pass? You've got a problem with, say, your AS chunk of routing space being full of IRC-controlled robot machines. So set your router to forward all IRC packets (in- or out-bound) to /dev/null (or a logging system) and then annihilate any IRC bot-controllers in your system.
If IRC has some value (which I have yet to be shown an argument for, let alone be convinced by such an argument ; "Look at this, it's kewl!" is not an argument), then tell the developers who claim so to come up with an IRC-like system which is provably secure and that provides the functionality they want without the security risks. Any of the security risks. Which returns to the original point - what is the "value" of IRC that people tolerate the security risks that appear to be inherent in the model.

Question: What did people do for rapid networked communication between self-selected groups before someone (whoever) invented IRC? Answer : mailing lists and/or private newsgroups on non-peering, non-usenet NNTP servers.
Question: What is still a major method of rapid networked communication amongst self-selected groups? Answer: mailing lists (and private newsgroups too, but often less visible than the lists). Did you notice that SourceForge provides this functionality? You think it's there to make the menus longer, or for some other reason?

If it causes pain, and you've got an alternative, stop doing it.

BTW, who was responsible for this junk? I remember something similar being available on Compuserve when I joined in 1992, but it was unusable then and hasn't got any better since.
It is possible that the security risks of IRC are consequent on the possibility of being anonymous on the communication system. That may account for a lot of the junk too. Although the IRC-like stuff in Compuserve was on a private network with personal accountability through credit-card-backed account identifiers, and that was pretty content-free.

Re:Why allow IRC?

I'm not going to argue the merits or faults of IRC, because it doesn't matter. The problem is that even if you say "IRC is bad" there isn't really a way to "not allow" it. Generally IRC uses ports around the range 6669-7000 (IIRC). So everyone firewalls those off... And the owners of the server move to port 3456 (or whatever...)

So just port filtering doesn't work. The next idea is to do stateful packet inspection. Every router looks at the contents of every packet to determine if it is part of the IRC protocol.

Ok, this would work, except it would be unacceptably expensive to implement. Plus, I beleive that some (most? all?) IRC servers support SSL and possibly IPSEC. So the packets are encrypted using SSL, and using some non-obvious port. (like say, port 443) At this point, it is very hard to distinguish between legitimate HTTPS traffic and IRC traffic. I suppose you could look at the packet sizes and do traffic analysis on the flows, but you'd still have problems with other legitimate services running over HTTPS. (Like VPN proxies or Java Applets, or Flash)

So, even if IRC is the root of all evil in the world, it's not possible to just "not allow" it.

(Sorry for the rant, I'm getting over being sick and still a bit punchy)

Will XOR be used for next generation botnets?

So if they are now able to shut down C&C , what will the next round of innovation bring to malware?

Will next generation botnets probably use XOR metrics to receive their instructions, similar to serverless p2p nets?

I think that's just a pyrrhic victory ... one more such victory and we are lost ... does it even need one more such victory?

nuking botnets

How about if you "poisoned" the pool of botnets. Since there has to be a master machine sending the "attack" signal, perhaps one could make them turn upon that master. That is to say, modify the botnet binaries running on infected machines and replace them with a modified version.

Next time the botnet master says 'attack IP 192.168.253.1' the botnets instead turn on whomever issues the attack command. This would likely be an "owned" machine and not the attacker's home, but at least it gets knocked off and disabled. Another idea might just be to have the machines redirect the floods to 127.0.0.1, thus incapacitating themselves instead.

Re:And then what?

Let the military handle that. They have nice tanks that can blow any infrastructure you tell them to.

Re:Who cares really

But it doesn't hurt anyone else much either as I'm on a 56k line. Oh, scary DOS comming from that.

What you're saying shows the root of the problem and why it's so hard to solve: you need some level of cooperation from people who do not have a direct interest in solving it simply because it doesn't affect them. Sure, your little 56k is quite harmless, but with 1000 zombies on little 56 lines, you can create quite a flood.

The other problem is with using up bandwidth allotments. Let's say the attacker is using 2KB/s for flooding. You won't notice that, but the other end wastes 5GB/month. Now if you have just 200 56k lines on pumping this on average, you'll be driving the target into unwanted bandwidth bills for sure. Now this analysis is making some assumptions, but you get the picture.

Re:Who cares really

This isn't flamebait, he's making a point.

Most 'normal' users really don't seem to give a damn if their computeris being hijacked, as long as they don't notice it. And the same users won't undertand that their 56k line is one of many, which adds up to an enormous amount of bandwidth.

Re:Who cares really

It's a zombie! Aim for the head!

Re:Ex-Soviet Hackers...

"Strike the shepard, and the sheep will scatter"

That's not true at all, sheep generally hang around in big mobs regardless of whether a shephard is present or not.

Even if what you said was true there is no evidence it would work for anything other than shephards and sheep.

Word on what street ?