Sunggwan Choi

Red team operator | Struggling to catch up, enjoying the struggle along the way. 특 해장국에 선지빼고 후루룩 하려고 노력하는 중

[Bad Sector Labs @badsectorlabs]

Babe, wake up. Someone finally dropped a sick macOS/Linux implant.

https://t.co/rMmAxoI2gr

[SpecterOps @SpecterOps]

Your browser is a C2 agent waiting to happen.

@KingOfTheNOPs shows how to silently sideload a Chromium extension, no prompts or GUI, and turn Chrome/Edge into a persistent implant with SOCKS, cookie theft, and OS access.

Read more 👇 https://t.co/kh0W1ObkeQ

[📔 Michael Grafnetter @MGrafnetter]

CVE-2026-34348 exploitation demo from my #BHUSA "Pass-the-Passkey Family of Attacks" talk: WebAuthn assertion from recent YubiKey authentication is extracted from Windows Event Log and replayed against Microsoft Entra ID using Passkey Injector.
Whitepaper: https://t.co/bp5sriMB5z

[David @dmcxblue]

Just in time for DEFCON/BlackHat ANIMO an Azure C2, sorry for the poor README but I will be sure to update in the coming days.
#redteam
https://t.co/1jI6iI3616

[Atsika @_atsika]

I've noticed that @bunjavascript has been gaining a lot of traction lately, so I thought it might be interesting to take a look at its offensive potential 🕵️

The game is evolving and we have now entered a new post-exploitation era made of Virtual Machines and Interpreters. Let's revive some old techniques using modern tools ♻️

https://t.co/4DUeJhMcVq

[nix @aniqfakhrul]

Happy to share a technical analysis by @h0j3n on our recent CVE-2026-54121 a.k.a Certighost. glhf🔥

Technical analysis: https://t.co/jYrZDLPlJr
POC: https://t.co/yBeHg1vQHP

[Florian Roth ⚡️ @cyb3rops]

One thing about this OpenAI / Hugging Face incident really bothers me. Hugging Face says the intrusion was driven “end to end” by an autonomous AI agent system.

But how do they actually know that?

Victim-side telemetry can show automation, speed, thousands of actions, short-lived sandboxes, changing infrastructure etc. It cannot show what happened upstream.

It cannot tell us whether humans changed prompts, restarted runs, selected successful paths, provided more context, redirected agents or manually helped at certain points. We also don’t know what was actually decided by a model and what was simply automated by the surrounding agent framework.

Maybe OpenAI has all those traces. Fine. Then publish them.

Show the prompts, tool calls, failed runs, model handoffs, restarts and human interventions. Without that, “end-to-end autonomous” is a claim, not a proven technical finding.

The forensic-refusal dataset Hugging Face published proves something much smaller: https://t.co/LemUxeaY4V

It shows that Claude refused to analyze one small Python backdoor while GLM 5.2 completed the analysis. That is a valid example of hosted-model guardrails getting in the way of incident response. But this is not evidence that the intrusion itself was carried out end to end by an autonomous agent.

And this claim matters because it pushes a very specific idea into people’s heads: AI agents can now independently find zero-days, escape sandboxes, move laterally, steal credentials and compromise companies.

Then comes the second part of the story: Hugging Face used local AI models to investigate the AI attacker “at machine speed”.

So the message basically becomes:

- AI attacked us
- AI helped save us
- Therefore, everyone needs more AI

Come on 🙄
Weak isolation, excessive privileges, poor credential boundaries, insufficient segmentation and far too much blast radius.

You don’t need an AI defender to fix those things.

Even fairly basic controls like rate limits and temporary blocks across source IPs, accounts, tokens and job volume could have throttled at least parts of this activity and created a very obvious signal for an analyst to review. Add proper egress restrictions, isolated workers and credentials that do not open the door to production clusters .. none of this requires an LLM

Using a local model to analyze 17,000 events may have helped during the investigation. Good - I’m not questioning that. But that happened after the compromise.

What I really hate is that something which would have been an embarrassment ten years ago is now repackaged as a capability demo, a heroic AI-vs-AI story and a marketing pitch.

Maybe the attack really was fully autonomous. Then show the evidence. Until then, I don’t think this claim should be repeated as if it had already been proven.

Sources
https://t.co/1yi9ck5xWD
https://t.co/TSlel0Cyfz
https://t.co/LemUxeaY4V

[Nick VanGilder @nickvangilder]

Red team and offensive security tradecraft is scattered across maybe a few hundred operator blogs. Some are active. Some are stale. It’s kind of hit or miss but definitely a PITA to keep track of them all tbh. Over the past few months, I’ve been working on identifying, cataloging, and indexing as many of them as possible in a centralized location, now housed here: https://t.co/XznrLlfkqo. By the numbers, there is currently 6,500+ unique tradecraft-specific blog posts by ~190 unique individuals (many of them operators, researchers, or practitioners) and growing daily.

You might notice that the search box has gotten a facelift and upgrade compared to other pages on the site. It’s a RAG pipeline that you can ask questions using natural language, and the retrieval is pretty cool too. I feel like a lot of “chat with the docs” tools embed everything into vectors and just call it a day. That ends up handling fuzzy questions moderately well but isn’t practical for what operators would actually search for.

Retrieval runs two ways at once. First, the question is rewritten into a keyword rich query. Then a dense retriever (vector embeddings) finds posts by meaning while a sparse retriever (postgres full text) matches exact terms literally, in parallel. The two rankings are then fused with RRF (reciprocal rank fusion), and an LLM reranks the survivors against the original question. The top posts go to the model, which answers strictly from them, with citations. If the corpus doesn’t cover it, it says so instead of inventing some made up hallucinated bs.

Basically, you can ask the way you think: “how do operators bypass AMSI to run .NET tooling in memory?” That should return specific writeups from multiple authors with each technique credited to the operator who published it. Or maybe you might ask, “what does modern phishing tradecraft look like in 2026?” That should surface evilginx style MFA bypass, Cloudflare Turnstile evasion, and primary refresh token phishing, with links to the sources. No hallucinated slop since it’s a RAG.

One design decision I care a lot about that I want to call out is attribution. So, even when the tradecraft originates from a company blog, the post is still credited to the individual who wrote it (not the company’s social account). This was intentional and a little extra work, but well worth it IMO. It’s not to take anything away from the teams at companies like SpecterOps or TrustedSec who put out amazing content on their company blogs. I love you guys! I really just wanted to try to focus on showcasing tradecraft authors and their personal contributions across the field.

Lastly, there is an entire discovery pipeline that will attempt to find new sources (new tradecraft authors) on a daily basis. The RAG also checks the sources for new posts on a daily basis too. That said, discovery isn’t perfect. So, if you know of a tradecraft blog that is missing, please consider adding it using the +add button on the site.

[klez @KlezVirus]

[TALK] My latest Black Hat Europe talk is now publicly available. If you can look past the painfully obvious anxiety and a speaker who occasionally sounds like his brain has stopped cooperating, you might find something useful. Who knows?

https://t.co/s1ms9xSB2I

[Adam Chester 🏴‍☠️ @_xpn_]

New blog post is up looking at how LLMs are making local EDR rulesets, YARA rules, and behavioral detections trivial to extract. This post focuses on how simple the harness can be. Buckle up h4xx0rs, the next few months are gonna get interesting! https://t.co/QvzXsPA01F

[Adam Chester 🏴‍☠️ @_xpn_]

First blog post in a mini series where I look at "disposable tooling". This post shares what I have found to be useful when 1-shot'ing LLM generated Stage-0 agents for Mythic. https://t.co/LX0JVGmHBf

[Praetorian @praetorianlabs]

AMSI and ETW are built to watch the .NET runtime. So we took the runtime away. ⚒️

WasmForge compiles Rubeus and Seatbelt to WebAssembly and runs them outside the CLR completely. Open source as of today.

https://t.co/vZmDbVY6nP

#OffensiveSecurity #Praetorian #PraetorianGuard

[Michael Weber @BouncyHat]

For folks looking to reproduce our Sliver compilation process, you can do this now following the directions at https://t.co/5exP43TGjt.

[drew @hoodoer @hoodoer]

I'm excited to get v3 of JS-Tap released publicly.

3 new beacons for persistence on endpoints, BEX (malicious browser extension), Atom (electron app implant), and V8 (node/bun app implant). Read more here:
https://t.co/7LphrdF0Fo

Find me at BlackHat Arsenal for a demo.

[Michael Weber @BouncyHat]

We think of WASM as a mechanism to run compiled code in your browser, but what if we shimmed in all the host APIs necessary to run full implants with ALL logic entirely in the WASM VM? This post walks through what that looks like.
https://t.co/xGVpPe2zyC

#wasm #malware #sliver

[Sunggwan Choi @_choisec]

@0xmh1

[Nick VanGilder @nickvangilder]

I haven’t been as active on the socials lately, because I’ve been working on a community project that’s kept me pretty busy. That said, I think I’m finally far enough along with it that I can share the project in its current state and talk more about it.

So, I present to you: https://t.co/EqYQFPuvrF

I bulit site this for a few reasons, but one of the main reasons was/is that I didn’t feel like there was a centralized resource for red teamers that included all the things that red teamers tend to care about. I also wanted to build something that the community could add to, edit, maintain, etc., while also being self-updating, self-healing, and less likely to go stale over time. So, there’s quite a few different cron jobs, GitHub actions, AI calls, API calls, and other workflows that trigger at set intervals and patterns to try to keep it fresh. For example, I’m leveraging various sources (e.g. conference websites) that help identify conference talks which then feeds into a YouTube API to identify conference talks based on certain criteria. I realize there’s still lots work to do, and I’m fully aware that this is a not a 100% fully functioning site at this time. If you have any ideas for improvements, want to report a bug, want to help be a maintainer, or really anything at all, just let me know. I welcome any and all feedback or help!

Also, I know there is a lot of interest in the Scenario Generator module (which I posted about a couple of weeks ago); however, I can't open source it at this time, and it's not currently operational due to Claude API costs to power it. I am still sorting through how to make this available to the community at no charge; however, it may not be possible for what it costs to produce output. More to come on this module! While I sort it out, I am also redesigning it, and you are welcome to check it out in its current state.

[Andy Grant @andywgrant]

Over time, I have developed a bit of a reputation for saying no to external work for my teams. At least that's how people describe it. But I realized what I actually do is shape the work, often before it even arrives.

https://t.co/x00SM6H1oX

[%TEMP% @TEMP43487580]

Thank you for coming to today's talk at #Insomnihack . I will upload the slides later but here is the new release of the BAADTokenBroker bof for those of you interested

https://t.co/NUxFLzY7hO

[klez @KlezVirus]

[RELEASE] Better late than never! Part 3 is out! Fantastic unwind information and where to find them. We went digging through .pdata, RTF Lookups, and a few ntdll internals that probably weren't meant to be touched. BYOUD dropping alongside. Enjoy 😉

https://t.co/gQ6lRFctO7