RedWarden - Flexible CobaltStrike Malleable Redirector

RedWarden - Flexible CobaltStrike Malleable Redirector

(previously known as proxy2's malleable_redirector plugin)

Let's raise the bar in C2 redirectors IR resiliency, shall we?

Red Teaming business has seen several different great ideas on how to combat incident responders and misdirect them while offering resistant C2 redirectors network at the same time.

This work combines many of those great ideas into a one, lightweight utility, mimicking Apache2 in it's roots of being a simple HTTP(S) reverse-proxy.

Combining Malleable C2 profiles understanding, knowledge of bad IP addresses pool and a flexibility of easily adding new inspection and misrouting logic - resulted in having a crafty repellent for IR inspections.

Should any invalid inbound packet reach RedWarden - you can redirect, reset or just proxy it away!

Abstract

This program acts as a HTTP/HTTPS reverse-proxy with several restrictions imposed upon inbound C2 HTTP requests selecting which packets to direct to the Teamserver and which to drop, similarly to the .htaccess file restrictions mandated in Apache2's mod_rewrite.

RedWarden was created to solve the problem of IR/AV/EDRs/Sandboxes evasion on the C2 redirector layer. It's intended to supersede classical Apache2 + mod_rewrite setups used for that purpose.

Features:

Malleable C2 Profile parser able to validate inbound HTTP/S requests strictly according to malleable's contract and drop outlaying packets in case of violation (Malleable Profiles 4.0+ with variants covered) Ability to unfilter/repair unexpected and unwanted HTTP headers added by interim systems such as proxies and caches (think CloudFlare) in order to conform to a valid Malleable contract. Integrated curated massive blacklist of IPv4 pools and ranges known to be associated with IT Security vendors Grepable output log entries (in both Apache2 combined access log and custom RedWarden formats) useful to track peer connectivity events/issues Ability to query connecting peer's IPv4 address against IP Geolocation/whois information and confront that with predefined regular expressions to rule out peers connecting outside of trusted organizations/countries/cities etc. Built-in Replay attacks mitigation enforced by logging accepted requests' MD5 hashsums into locally stored SQLite database and preventing requests previously accepted. Allows to define ProxyPass statemtents to pass requests matching specific URL onto other Hosts Support for multiple Teamservers Support for many reverse-proxying Hosts/redirection sites giving in a randomized order - which lets load-balance traffic or build more versatile infrastructures Can repair HTTP packets according to expected malleable contract in case some of the headers were corrupted in traffic Sleepless nights spent on troubleshooting "why my Beacon doesn't work over CloudFlare/CDN/Domain Fronting" are over now thanks to detailed verbose HTTP(S) requests/responses logs

The RedWarden takes Malleable C2 profile and teamserver's hostname:port on its input. It then parses supplied malleable profile sections to understand the contract and pass through only those inbound requests that satisfied it while misdirecting others.

Sections such as http-stager, http-get, http-post and their corresponding uris, headers, prepend/append patterns, User-Agent are all used to distinguish between legitimate beacon's request and unrelated Internet noise or IR/AV/EDRs out of bound packets.

The program benefits from the marvelous known bad IP ranges coming from: curi0usJack and the others: https://gist.github.com/curi0usJack/971385e8334e189d93a6cb4671238b10

Using an IP addresses blacklisting along with known bad keywords lookup through Reverse-IP DNS queries and HTTP headers inspection, brings the reliability to considerably increase redirector's resiliency to the unauthorized peers wanting to examine attacker infrastructures.

Invalid packets may be misrouted according to three strategies:

redirect: Simply redirect peer to another websites, such as Rick Roll. reset: Kill TCP connection straightaway. proxy: Fetch a response from another website, to mimic cloned/hijacked website as closely as possible.

This configuration is mandated in configuration file:

Below example shows outcome of redirect to https://googole.com:

Use wisely, stay safe.

Requirements

This program can run only on Linux systems as it uses fork to spawn multiple processes.

Also, the openssl system command is expected to be installed as it is used to generate SSL certificates.

Finally, install all of the Python3 PIP requirements easily with:

Usage

Example usage

The minimal RedWarden's config.yaml configuration file could contain:

Then, the program can be launched by giving it a path to the config file:

The above output contains a line pointing out that there has been an unauthorized, not compliant with our C2 profile inbound request, which got dropped due to incompatible User-Agent string presented:

Use Cases

Impose IP Geolocation on your Beacon traffic originators

You've done your Pre-Phish and OSINT very well. You now know where your targets live and have some clues where traffic should be originating from, or at least how to detect completely auxiliary traffic. How to impose IP Geolocation on Beacon requests on a redirector?

RedWarden comes at help!

Let's say, you want only to accept traffic originating from Poland, Europe. Your Pre-Phish/OSINT results indicate that:

89.64.64.150 is a legitimate IP of one of your targets, originating from Poland 59.99.140.76 whereas this one is not and it reached your systems as a regular Internet noise packet.

You can use RedWarden's utility lib/ipLookupHelper.py to collect IP Geo metadata about these two addresses:

The former brings:

and the latter gives:

Now you see that the former one had "country": "Poland" whereas the latter "country": "India". With that knowledge we are ready to devise our constraints in form of a hefty YAML dictionary:

Each of that dictionary's entries accept regular expression to be matched upon determined IP Geo metadata of inbound peer's IP address. We use three entries in country property to allow requests having one of specified values.

Having that set in your configuration, you can verify whether another IP address would get passed through RedWarden's IP Geolocation discriminator or not with ipLookupHelper utility accepting second parameter:

The very last line tells you whether packet would be blocked or accepted.

And that's all! Configure your IP Geolocation constraints wisely and safely, carefully inspect RedWarden logs for any IP Geo-related DROP entries and keep your C2 traffic nice and tidy!

Repair tampered Beacon requests

If you happen to use interim systems such as AWS Lambda or CloudFlare as your Domain Fronting / redirectors, you have surely came across a situation where some of your packets couldn't get accepted by the Teamserver as they deviated from the agreed malleable contract. Was it a tampered or removed HTTP header, reordered cookies or anything else - I bet that wasted plenty hours of your life.

To combat C2 channels setup process issues and interim systems tamperings, RedWarden offers functionality to repair Beacon packets.

It does so by checking what Malleable Profile expects packet to be and can restore configured HTTP headers to their agreed values according to the profile's requirements.

Consider following simple profile:

You see this Accept-Encoding? Every Beacon request has to come up with that Header and that value. What happens if your Beacon hits CloudFlare systems and they emit a request that will be stripped from that Header or will have Accept-Encoding: gzip instead? Teamserver will drop the request on the spot.

By setting this header in RedWarden configuration section dubbed protect_these_headers_from_tampering you can safe your connection.:

Example outputs

Let's take a look at the output the proxy produces.

Under verbose: True option, the verbosity will be set to INFO at most telling accepted requests from dropped ones.

The request may be accepted if it confronted to all of the criterias configured in RedWarden's configuration file. Such a situation will be followed with [ALLOW, ...] entry log:

Should the request fail any of the checks RedWarden carries on each request, the corresponding [DROP, ...] line will be emitted containing information about the drop reason.:

Drop Policies Fine-Tuning

There are plenty of reasons dictating whether request can be dropped. Each of these checks can be independently turned on and off according to requirements or in a process of fine-tuning or erroneus decision fixing:

Excerpt from example-config.yaml:

By default all of these checks are enforced.

Turning debug: True will swamp your console buffer with plenty of log lines describing each step RedWarden takes in its complex decisioning process. If you want to see your requests and responses full bodies - set debug and trace to true and get buried in logging burden!

FAQ

- Can this program run without Malleable Profile?

Yes it can. However request inspection logic will be turned off, the rest should work fine: IP Geolocation enforcement, reverse-lookup logic, banned IPs list, etc.

- Can this program be easily adapted to other C2 frameworks as well? Like Mythic, Covenant, etc?

Easily no. With some efforts - yes. As I've described below, the tool is written badly that will make other C2s adaptation a pain. However that's totally doable given some time and effort.

- My packets are getting dropped. Why?

Try to enable debug: True and trace: True to collect as many log as possible. Then you would need to go through logs and inspect what's going on. Do the packets look exactly how you expected them in your Malleable profile? Or maybe there was a subtle tamperation along the network that causes RedWarden to drop the packet (and it could make Teamserver drop it as well?).

Known Issues

It may add a slight overhead to the interactive sleep throughput ProxyPass processing logic is far from perfect and is really buggy (and oh boy its ugly!). Weird forms of configuration files can derail RedWarden parser and make it complain. Easiest approach to overcome this would be to copy example-config.yaml and work on it instead.

Oh my god why is this code such an engineerical piece of crap?

The code is ONE FUCKING BIG HELL OF A MESS - I admit that - and there's an honest reason for that too: the project was being developed 90% during the actual Red Team engagements. As we all know, these sorts of engagements entail so many things to do, leaving close to no time for a proper complex tool development. Not to mention criticality of this program in project's setup. The tool initialy started as a simple proxy script written in Python2, to then evolve as a proxy with plugins, received malleable_redirector plugin - and since then I've been trying really hard to keep proxy2 maintain backwards compatibility (poor me, I was like Microsoft!) with other plugins I've made for it and stick to its original purpose.

Time has come though to let it go, rebrand it and start fixing all the bad code smells introduced.

With all that said, please do express some level of compassion for me when raising issues, submitting pull requests and try to help rather than judge! :-) Thanks!

TODO

Research possibility to use Threat Intelligence feeds to nefarious purposes - like for instance detecting Security Vendors based on IPs Add support for MaxMind GeoIP database/API Implement support for JA3 signatures in both detection & blocking and impersonation to fake nginx/Apache2/custom setups. Add some unique beacons tracking logic to offer flexilibity of refusing staging and communication processes at the proxy's own discretion Introduce day of time constraint when offering redirection capabilities (proxy only during office hours) Add Proxy authentication and authorization logic on CONNECT/relay. Add Mobile users targeted redirection Add configuration options to define custom HTTP headers to be injected, or ones to be removed Add configuration options to require specific HTTP headers to be present in requests passing ProxyPass criteria. Interactive interface allowing to type simple characters controlling output logging verbosity, similarly to Nmap's Rewrite Malleable profile parser logic to pyMalleableC2. When I first started coding my own parser logic, there was no such toolkit on Github. Refactor all the codebase

Author