Thread by @hacks4pancakes

Sometimes instead of blogging I feel like making a big old Twitter thread, so let's talk about Cobalt Strike for people only vaguely familiar (or misinformed) with the concept. Maybe I'll blog it later.

Cobalt Strike is an adversary enumeration tool used to train teams how to do incident response and threat hunting. It was made by a genius I genuinely like and will not disparage, Raphael Mudge. The first time I met him he flew across the floor air-guitaring in his dress clothes.

A lot of you are familiar with the easy-button hacking tool, Metasploit. Well, he made this shnazzy GUI for Metasploit called Armitage.

But, he realized it was still tough for a lot of defenders to get highly skilled Red Teams to train them. Or sommat, I'm not in his head...

Ask him for his take.

Anyway, our pal made this genuinely incredible software suite called Cobalt Strike that not only does payloads & exploitation + shells, but also allows for team-based command and control and sophisticated scripting. Distributed hacking.

This means that a low-moderately skilled Red Team can run a really impressive, realistic intrusion scenario as a coordinated group with just a few experts to help out as needed.

All the C2 beacons are tidily managed and centralized. C2 profiles can be emulated. It's amazing.

Okay but here's the problem. Cobalt Strike works ... really, really well. It works so well at emulating adversaries that bad guys were like:

"Yo, this would make our skids' lives easier..."

So their real experts worked overtime to crack the (genuinely well-controlled) tool.

So now, for the last few years, we've seen Cobalt Strike pop up all over in real intrusions. It's really good at what it does, and it is tremendously customizable.

People and tools will tell you they can detect it but they just detect pieces, behaviors, and defaults.

So, that sounds really daunting. Like, how do you detect this really sophisticated tool? Well, part of it is understanding how it works. It's still facilitating an intrusion that traverses layers of the kill chain. There are good opportunities to catch it.

Cobalt Strike relies heavily on C2 beaconing. Yes, you can build fancy profiles that look like Amazon or Pandora. Yes, it can beacon at irregular intervals. However, C2 rules still apply, and a lot of adversaries use the defaults. Detecting HTTP, HTTPS, or DNS C2, that's a start.

Cobalt Strike has to be delivered to a system somehow. Yes, that could be hack-fu, but the defaults are old but effective nonsense like phishing with malicious links to stuff like HTAs. Always panic when you see HTAs, and do good phishing and watering hole detection.

Cobalt Strike can be memory-resident only when injected into a process. A novice may create a new process which will look quite anomalous in EDR, and can be killed. A more experienced person will rapidly migrate into a system process you can't kill. But it won't survive shutdown.

Volatility and EDR are solid bets to detect injected processes. But did you know there are several public PowerShell scripts which can find and carve injected process space, effectively killing Cobalt Strike? Not a promising tool for real life, but a fun, fun, fun one for games.

Cobalt Strike can, of course, establish persistence. Yes, there are a billion ways to do this, but a really simple easy button to click is just making a new auto-starting service. Always check your service creation and abnormalities. Sysmon is your friend. Use @SwiftOnSecurity's

Cobalt Strike can also abuse temporary services just to start itself up in a privileged process. Always be on the lookout for high entropy and unknown services starting then vanishing. It may also drop, then promptly delete high entropy executables.

Once credentials have been stolen from a system (or if there's credential reuse) an adversary can use Cobalt Strike to move laterally to one system to another without establishing internet beacons directly to it. Sneaky! It uses the first system as a relay.

That's not the worst case ever, because the default methods to do this tend to be familiar stuff like WMI or PS abuse. Cobalt Strike may also establish named pipes between systems. It has four default names. You can look at named pipes with many tools like OSQuery and PowerShell.

So, a couple places you can catch that lateral movement - at the host, and at the network. Unencrypted exfil of Stuff Cobalt Strike Encourages Stealing is also quite visible at the network. For instance, credential dumps. Look for contents and sizes.

So a summary:
- The creator of Cobalt Strike is a good dude!
- It seriously sucks it's been stolen and repurposed
- It's a cool framework!
- It's a huge sandbox and nobody can claim to detect every configuration,
- Some defaults and behaviors, you can catch.

- Coverage across the kill chain, host and network, is really important in catching more sophisticated attacks like this
- If you catch one small component of Cobalt Strike, like lateral movement or an injected process, or beacons using it's default profiles, pull the thread.

- PowerShell and Sysmon are underrated host analysis tools, even if you don't have the most advanced EDR deployed widely. They take care and feeding.
- PowerShell is just as powerful for adversaries. Log PowerShell. Everyone freaking abuses it.
- Network monitoring is a must.

- Consider dedicated threat hunt hypotheses for Cobalt Strike TTPs in your environment. It's not just a cheaty way to catch the red team anymore. It's really being abused by a lot of bad guys, often times very clumsily.

I first started using Cobalt Strike in 2015, and it has been incredible to see its capabilities and the libraries of Aggressor Scripts and C2 Profiles for it grow. If you have the opportunity to get hands on with it I highly recommend it. It can be tough and $$$ to get a license.

However, it is really invaluable as a defender to understand at some level how it works and how to manipulate it - and what the default "easy button" options are.