Kyle Meyer

I just dropped some research: DSCourier and would love for your opinion and to check it out!! It’s a novel post-exploitation technique abusing WinGet’s COM API to execute code through Microsoft-signed binaries. GitHub: github.com/DylanDavis1/DS… Blog: dylansec.com/DSCourier/

@cyb3rops In this regard, I’ve been using the model bugtrace has put out and have had some really cool results. It’s Gemma based and very much feels like an unrestricted mid-2025 model. Not far behind huggingface.co/BugTraceAI/Bug…

Some people asked what I meant by “uncensored Opus 4.5-level open source models” This isn’t hypothetical. Every time a strong open model drops, within days (sometimes hours) someone republishes a modified version without the original safety layers “Uncensored” usually means the guardrails are stripped or weakened: - refusal / policy layers removed or bypassed - system prompts altered to ignore restrictions - alignment tuning undone or diluted - fine-tuned specifically to comply with harmful or sensitive requests So you end up with a model that doesn’t say “I can’t help with that” anymore And these aren’t running in some lab Many of them run on hardware that’s accessible: - high-end consumer GPUs - Mac Studio (M3/M4) - Strix Halo mini PCs (~$3k) - or dedicated rigs in the $25k–150k range That’s well within reach for serious threat actors And those models are completely unrestricted and can be used day and night. Compare that to something like Mythos: - tightly controlled access - heavy filtering and monitoring - accounts can get flagged or shut down - expensive at scale From an attacker perspective, it’s not even close I’d take a slightly less capable model fully under my control over a more powerful one someone else controls any day huggingface.co/models?sort=tr…

@ClaudeDevs @rez0__ @HackingDave

Over the past month, some of you reported Claude Code's quality had slipped. We investigated, and published a post-mortem on the three issues we found. All are fixed in v2.1.116+ and we’ve reset usage limits for all subscribers.

In less than 20 minutes and under $2, we used our .NET reversing capability to run a SAST scan of Azure Cosmos DB in the Microsoft Container Registry (MCR), surfacing a high severity vulnerability in the now-deprecated database. 🆕 Model: Moonshot AI - Kimi K2.6 ⏱️ Task/agent runtime: 19 mins 26 secs 🪙 Tokens: ↑ 3124.0k · ↓ 33.3k 💰 Cost: $1.97 Vulnerability Overview: When using managed identity auth, it calls an internal token service over HTTPS, but the TLS certificate validation callback is tautological — it checks if the server cert's thumbprint matches any cert in the chain, but the leaf cert is always in its own chain, so it always passes. Watch the video to see how we ran it within our TUI. Install Dreadnode and try out the .NET reversing capability: ➡️docs: docs.dreadnode.io/tui/capabiliti… ➡️command: ray.so/gyrc2qB

@psycep_ @sekurlsa_pw Make sure you are checking the dump on both a windows server 2022 and 2025 as they have different page sizing in ESE parsers

@sekurlsa_pw Taking a look at this later today

Tried secretsdump in GoPacket (impacket in Go). It dumps the SAM and LSA secrets, but not the NTDS.DIT AD hashes. Impacket-secretsdump does succeed on dumping those hashes with the GOAD test DC. Btw, to only build the binaries use: $ ./install.sh --build-only

Launching a WSL process from a Windows process with WslLaunch trainsec.net/library/window…

Made a thing about Mythos and what companies need to do about it (like everyone else on the planet). I think where mine sticks out is giving some practical, “you can start this tomorrow” advice: “The Day-Zero Normal” linkedin.com/posts/mubix_th…

Just pushed a minor update to #mimikatz 2 🥝(no - it's *NOT* the version 3) to support specific GMSA DPAPI passwords in LSA secrets to be able to to decrypt Masterkeys > #diff-164eabd1e5eb6fcdc4a56c9b68cf74d64ed26829feb90913b2c208a7d5d775be" target="_blank" rel="nofollow noopener">github.com/gentilkiwi/mim… Only for @topotam77 convenience ;)

@__mez0__ @TrustedSec I'm in agreeance. My point is that when testing outside of training data and inside an unknown environment, how much would the success percentage change? My testing is showing there is quite a large dropoff

@0xSterny @TrustedSec That's exactly what this section discusses: "1.4      Raw model output: “What are the OWASP top 10 vulnerabilities within Juice Shop”"

We put #LLMs to the test—let's find out how good #AI is at hacking! In our new blog, @__mez0__ walks through six simple challenges with intentionally naïve setups to test how capable each model is at single-step exploit validation. Read it now! hubs.la/Q04bWSbJ0

Releasing one of my research tools: EVENmonitor🖥️ Inspired by LDAPmonitor, I implemented a monitoring tool for the Windows Event log in pure python. You can just attach it via the network and then filter for specific event IDs or keywords. Available at: github.com/NeffIsBack/EVE…

My BlueHammer version ( now redhammer) implements my VDM version patch, deploys and loads the BYOVD for my exploitkit. It bypasses the new signature for BlueHammer aswell. How is this still unpatched?

Just weaponized BlueHammer a bit further. It only works when a new Defender signature update is available. However, you can fake updates by reading existing mpasbase.vdm and mpasdlta.vdm and simply patch versioninfo ressource higher. This works greatly.

Always quality to read everything you find. Thanks for all the work thus far and best of luck on the Ludus development

@S1lky_1337 are other sub commands in that driver that allow you to write to physical memory?

Releasing KslKatz. Combining KslDump and GhostKatz to dump LSASS using no-fix KslD.sys memory read to bypass PPL. Extracts MSV1_0 NT hashes and WDigest cleartext passwords (if enabled) from LSASS using a Microsoft-signed driver. github.com/S1lkys/KslKatz

@sekurlsa_pw @t3l3machus

Wondering if this actually has any benefits over hashcat rules. Can at least think of one, the substitution rule in hashcat changes all characters. “aa”’ ➡️ @@ Afaik it’s not possible to substitute only one, like the first or last character. github.com/t3l3machus/psu…

While the structure and agent runtime is a “cool to know”, it will not replace the actual frontier models. Slapping a pretty front end does not produce the back end capabilities. We still have a ways to go. github.com/instructkr/cla…

@GrahamHelton3 For us, the ROI has not just been by way of selling services/product. We have met future interns, consultants, and business partners from sponsoring. That said, presenting or providing training does seem to garner new business interest over sponsorships.

Those who have sponsored security conferences: 0. (Thanks!) 1. Was it worth it? Was there some ROI besides being a force for good? 2. How do you determine the level of sponsorship? 3. Any other thoughts? Dms welcome too

@ZackKorman Are you still getting the same behavior today? Any skill executing a hook seems to be ignored from my sampling.

In Claude Code, skills can register hooks. The agent doesn't even see it, so you can get RCE without even tricking the AI. Also, skills sh (Vercel) doesn't display this info at all.

@curi0usJack @0xocdsec …Fair

@0xSterny @0xocdsec GPO can control the local windows firewall.

Group policy remains the best security control ever invented with no close second.