Digital Shokunin

Machine Learning

Everyone is into LLMs right now, ~48 billion of investment went into AI/ML in 2023 alone because of the hype surrounding ChatGPT and similar products. Even more investment going into it this year if the first quarter is any indication. AI/ML is important, even if you’re tired of the snake oil you were sold in the past on ML/AI and were tired of hearing about it before the hype cycle hit, if you’re in the security space at all, you should care because your business or customers care and they’re using it, spending large amounts of money on it and its directly touching your most valuable data (because that’s what the business is training them on). Just when I thought Nvidia stock/profits couldn’t go any higher after increased demand for GPUs because of the crypto hype cycle which was somewhat waning, the AI hype cycle hits and now Nvidia can print money as fast as they can ship silicon, with no evidence of supply meeting demand any time soon.

This post is to follow up some of the technical details for the talk I gave at the 2024 Red Team Summit. The talk itself covered the use of container registries and infiltration through CI/CD pipelines as a means of initial access and persistence. This post will cover some of the technical details and examples that I used in the talk.

The first thing to discuss is gaining initial access to a container registry. This can be done in a number of ways, but the most common is through the use of weak or leaked credentials. Once access is gained, the attacker can then upload a malicious image to the registry. This image can be used to gain access to the CI/CD pipeline and then to the production environment.

CI/CD pipelines exist in just about every company that does some kind of development. Some companies have more mature pipelines than others, but the rule still holds. Most companies with in house development have a Jenkins instance or some similar build orchestration software, code repositories, path to deployment, etc. Depending on your background you may be familiar with these environments to varying degrees. I started out doing somethin between running application servers and devops, maintaining deployment pipelines and helping developers troubleshoot issues. I even wrote my own share of code and deployed my own changes to a minor application that eventually went into production. This was done as an opportuntiy to get me more familiar with that side of things and I found the experience very valuable. Those experiences became invaluable when I went into red teaming and security research as I already had a baseline familiarity with these environments and how they worked, and more importantly, insight in how these environments could be exploited by an attacker.

Now that Microsoft is blocking macros for internet and externally sourced documents, I feel its safer to talk about some of the EDR evading Word macro techniques I have used in the past. Particularly for delivering Cobalt Strike beacons.

Cactus Torch is a great tool as a starting point. It takes some basic concepts such as a encoding the CS payload in memory, starting a process and injecting the payload into memory. Unfortunately, Cactus Torch is heavily signatured, but with a bit of modification, you can easily bypass most EDR solutions, including Windows Defender. Cactus Torch gets flagged because of a few function calls and variable names, but if you change those, you’ll find EDR no longer detects it. That’s the first step towards EDR evasion.

A while ago I earned my OSCP certification. Before that I had my GPEN and Pentest+. The Pentest+ I obtained during the beta program for the certification since the test was only $50 and I figured there was not much harm in trying. I took it practically blind (no preparation), and found out I passed in August. Shortly after I was given the opportunity to take the SpectreOps Red Team Training and after that scheduled to take OSCP training.

Back when CompTIA had a temporary beta program for the Pentest+, I took advantage of it, and the cheap cost of the test just to give it a shot. I didn’t study for it and kind of went in blind since no study materials existed and most of what was out there for it was pure speculation. It took a while to learn the results but I’m happy to report that I passed.

I’ve recently been using radare2 for a bit of reverse engineering and have used it a little bit in the past for CTF competions. (Side note: scaleway.com is a great cloud/VPS service if you need an ARM based server/machine for a something like a CTF to analyze ARM binaries and do not have a Raspberry Pi, ODroid or similar ARM based computer handy.)

I discovered Cutter recently, which has some instructions to compile it using cmake (also qmake but I had and used cmake).

2018 came with a big transition for me, a new opportunity came up that allowed me to pivot into InfoSec full time. Without going into too much detail I’ll be doing a bit of offensive security. This is very exciting for me to say the least.

One of the things I am working on that I have had plans to build out eventually for some time is a home pentesting lab. I’ve managed to setup a multi-core CPU PC with about 32 GB of RAM. I’ve setup a virtual environment using KVM that I plan to be managing with RHEL based tools. For now I’m using the Gnome Boxes and/or virt-manager. My plan is to simulate a fully functional environment networked VM’s to act as an attack range for both known and new exploits that I’ll use for practice or research.

The last year I’ve been busy working quite a bit with AWS. I’ve gone so far as to get my certification as the Associate Level as an AWS Architect. I accomplished this earlier in the year as well as renewing my Puppet Certification just recently for 2017. I may also be getting Docker certified.

Also related to AWS, I’ve recently discovered some infosec related resources on AWS from the offensive perspective. The first is Daniel Grzelak who has a blog on some of the possible way someone can back door an AWS account using AWS services. Another is a tookit called CS Suite which I haven’t had a chance to use extensively yet, but can help with auditing AWS security. AWS also provides some tools to do the same.

Atom.io was getting too bloated. So after much trial and error, and learning curve, I went hard into vim, and started using spacevim. In fact, I’m typing this up in spacevim right now with a live Markdown preview showing in a side browser window.

The biggest challenge besides getting it working and all the dependencies, was to get the terminal colors and unicode characters to display properly. At work, I settled on a Terminator terminal over X11+SSH session to get the colors just right, even though setting xterm-256color (or color256 maybe) as your terminal type and also installing the powerline fonts. The Deja Vu Mono from powerline works best I’ve found. Weirdly, it works better over an X11 session using a X Server on Windows (MobaXTerm) with vim running on a private cloud provisioned Ubuntu 16.04 server then it does on my Ubuntu workstation/laptop at home. The difference really is that just a few characters show up as random unicode blocks, even after changing the font on my home machine, not sure the difference (edit: I since fixed this by installing nerd fonts: https://github.com/SpaceVim/SpaceVim/issues/619 ).