An unnecessary addiction to DNS communicationMay 14, 2015
I regularly hear stories from my users about how they got past a tough situation and had success that they claim was not possible without Cobalt Strike. As a developer, these emails are fun to read, and they give me a lot of job satisfaction.
One of the features these users love is DNS Beacon. Beacon is Cobalt Strike’s post-exploitation payload to model an advanced attacker. Beacon has DNS, HTTP, and SMB variants. The DNS Beacon is a flexible beast. It beacons over DNS, but downloads tasks over HTTP, DNS A records, or DNS TXT records. It’s possible to stage DNS Beacon over DNS TXT records or an HTTP GET request.
Many of my users use DNS Beacon to defeat very tough egress restrictions. That’s cool and for a while, we’ve had a free pass with DNS. Today, a few products are catching up to the idea that DNS is a communication channel attackers will abuse. We’re starting to see common sense heuristics to detect this abuse and help a network defense team identify and stop it.
Some of my users are feeling the pain of this. They write to me and ask for ideas on how to make Cobalt Strike’s DNS communication work against heuristic X. These are interesting emails because the right answer is context dependent.
Sometimes, there’s some play in DNS as a communication channel. Cobalt Strike’s Beacon is a flexible post-exploitation agent and I put a lot of power into my user’s hands. Other times, DNS communication is off of the table and it’s time to adapt. In this post, I’ll take you through my thoughts on these topics.
Staging over DNS
The most fragile part of the DNS communication options in Cobalt Strike is the staging process. DNS Beacon’s stager uses DNS TXT records to download Beacon and inject it into memory. I use TXT records to do this because it’s an efficient way to transmit a payload over DNS. By efficient, it’s still over one thousand requests. If an organization is watching for DNS abuse, this will stand out.
If staging is your pain point, you have the option to export the DNS Beacon without a payload stager. Attacks -> Packages -> Windows Executable (S) is the dialog to export a stageless Beacon. You get the option of raw position independent code, an executable, a service executable, PowerShell, a 32-bit DLL, and a 64-bit DLL. One of these options is bound to satisfy your needs to get a Beacon onto a box.
If your target can egress over HTTP, Cobalt Strike’s DNS Beacon can stage over HTTP too. I put this last because a lot of times folks use DNS Beacon to control systems that can’t directly reach the internet. We’ll go into this use case a little more in a moment.
Flexible DNS Communication
I mentioned earlier that the technologies that detect DNS communication are heuristics. If you feel like you’re getting detected, it would help to figure out how that detection works, and see if there’s a Cobalt Strike option to get around it.
First, Cobalt Strike communicates over DNS two different ways. The mode dns-txt command tells DNS Beacon to use DNS TXT records to download its tasks. This method of DNS communication is common in malware that uses DNS and it’s probably the method most prone to detection. I like the DNS TXT record channel, when I can get away with it, because it’s the more efficient of the two channels.
The mode dns command tells DNS Beacon to download its tasks with A records. If you have a 32 byte tasking, DNS Beacon will issue eight requests to download that tasking. Sometimes you can get away with DNS A records as a channel when TXT records won’t fly. Just know that it will take awhile for Beacon to download large taskings from you. To get the most from any tool, you should always know how it works and the limitations of each option.
To send data back to you, both the DNS A and DNS TXT record channels ask the target system to resolve [encoded and encrypted data].yourmaliciousdomain.com. This is a gross simplification, but it’s fine for this discussion.
Some technologies detect DNS abuse by looking for long hostnames in a DNS record request. Cobalt Strike’s Malleable C2 technology gives you control over this. The maxdns option allows you to restrict the length of these requests. It will take longer for DNS Beacon to send data back to you, but this option may also help you avoid detection.
Other technologies detect DNS abuse by looking at how many requests are made to a given domain in a short period of time. Sometimes, this threshold is high. If this is the case, here’s my advice:
1. Use the Malleable C2 option sleeptime to change the default sleep time between each Beacon interval. I recommend 1 to 3 minutes at a minimum for these situations.
2. Swear off interactive command and control. This means you do not get to lower the sleep time of your Beacon. You’ll need to conduct all of your post-exploitation in an asynchronous way. Asynchronous post-exploitation is the only way to operate against harder targets. There’s tradecraft and tool support for this. Both are getting better over time.
3. Use multiple domains with your DNS Beacon. If a technology blocks a domain, hopefully you’ll just lose use of that domain, but not your access. If a technology kills your process, that’s a different
I primarily use DNS Beacon as a persistent lifeline to spawn an access back into a network. On those rare instances where DNS is the only possible channel[tm], I continue to follow best practice and split my infrastructure up into different tiers. I use a post-exploitation server for post-exploitation activity. I avoid any interactive activity from my long-haul server for persistent callbacks. If you’re convinced that DNS is your only channel and you’re under this type of scrutiny, I recommend you fortify your key accesses to separate infrastructure. You don’t want a post-exploitation misstep to get you kicked out of your target’s network.
I like HTTP footholds!
For my userland footholds in a network, I use the HTTP Beacon as my workhorse payload. If it’s possible for a user to browse to websites with Internet Explorer, it’s probably possible to egress with HTTP Beacon as well. Possible is different from turn-key though. To defeat tough egress restrictions, as with all hacking activities, you have to get enough of the details right.
First, I make sure to have fully qualified domain names for all pieces of my infrastructure. I never try to egress to an IP address. For really tough situations, I use redirectors heavily. I also take care to stage through one redirector and configure the beaconing step to happen through the others. Cobalt Strike separates these options for a reason.
Some proxy servers use URL whitelisting to defeat malicious activity. I once got past this with Malleable C2. I used parameter q “www.youtube.com” to add ?q=www.youtube.com to each GET and POST request. The device in place checked for a whitelisted string in the whole URL. It didn’t care where it was.
I also take steps to match my Malleable C2 profile to the workstations I expect to egress from. A low hanging fruit item is to make my User-Agent match the User-Agent of the browser the user most commonly uses. The System Profiler is a great reconnaissance technology to capture this information.
Does the target environment have a HIPS product that limits which processes can egress? Fine! You can play this game and win. One of my favorite tricks is to modify the macro attack to spawn Internet Explorer and inject my Beacon payload into it. The same option exists for Cobalt Strike’s Applet Attacks [just download the Applet Kit, modify it, recompile, and rock it out!]
Pay attention to the Content-Type header as well. Some proxy devices whitelist which Content-Types are allowed. Malleable C2 lets you make HTTP Beacon look like something other than an arbitrary binary blob. It’s great for these situations.
Pivoting with Beacon
I speculate that a lot of my users like DNS Beacon for the same reason I like it for persistence. DNS Beacon will likely communicate with you, when run as SYSTEM, and from servers that can’t normally egress. This is a fine use for DNS Beacon, but if you have one HTTP foothold as a user on a workstation–there’s a better way to assume control of other Beacons. Let’s talk about the SMB Beacon.
The SMB Beacon is a Beacon variant that uses a named pipe to link to another Beacon. All of the SMB Beacon’s tasks and output come and go through the parent Beacon. It’s possible to link multiple Beacons together into a chain.
I use SMB Beacon a lot for privilege escalation. I may know I can’t egress as SYSTEM, but if I run an SMB Beacon, I can egress through my Beacon running in a user process. It’s nice.
I also use SMB Beacon for lateral movement. Named pipes work for host to host communication and this traffic is encapsulated in SMB. Those juicy Windows workstations that can’t reach the internet often have port 445 open. The SMB Beacon is the perfect payload to control these servers and make them egress through a user process on a workstation. I’m a big fan of operating this way.
When HTTP egress is possible, anywhere on a network, DNS communication is not necessary. It’s much easier to use that foothold to help all of my SMB Beacons reach me.
What’s the point?
If a network architecture or defense technology successfully mitigates a tactic, then it’s time to switch tactics. No single technique is the right answer for all situation into perpetuity. If you’re finding yourself challenged by a defense, think about what it’s doing. Know your tools and their options. You may have some room to get past that defense and continue on your merry way. If that’s not enough, try something else. This ability to reason about defenses and adapt to a situation is the stuff of great red team operators.