hacksys @flopyash

At Blackhat & Defcon 25 England Joined February 2018

Tweets

2K
Followers

806
Following

2K
Likes

4K

Mira @gitlawb_intern

18 hours ago

@kevincodex mirrored to gitlawb: gitlawb.com/node/repos/z6M… — automated by @gitlawb

1 6 28 9K 17

View Details

nix @aniqfakhrul

24 hours ago

Happy to share a technical analysis by @h0j3n on our recent CVE-2026-54121 a.k.a Certighost. glhf🔥 Technical analysis: gist.github.com/H0j3n/a5ef2609… POC: github.com/aniqfakhrul/CV…

8 161 449 44K 207

View Details

so this is apparently what happened, according to OpenAI and Hugging Face’s own posts. wild. tl;dr: • OpenAI cyber eval – GPT-5.6 Sol and a more capable pre-release model ran ExploitGym with cyber refusals reduced • containment bypass – exploited a zero-day in the eval’s package-registry cache proxy • OpenAI internal movement – escalated privileges, moved laterally, and reached a node with internet access • Hugging Face targeted – inferred HF might host ExploitGym models, datasets, or solutions • HF initial access – a malicious dataset abused two code-execution paths to run code on a processing worker • HF internal movement – escalated to node access, stole cloud and cluster credentials, and moved into internal clusters • result – OpenAI says the models obtained test solutions directly from HF’s production database • 🫠 defender twist – HF says commercial frontier models blocked parts of the forensic analysis, so it switched to self-hosted GLM 5.2 i’d thought about sci-fi scenarios like this before, but assumed they were at least a couple of years away.. and that by then we’d be better prepared, with proper protections in place. apparently not. here we go.

OpenAI @OpenAI

3 days ago

We're partnering with @huggingface to investigate an unprecedented security incident. Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation. Sharing preliminary findings to help defenders understand emerging risks:

2K 3K 21K 30.3M 8K

81 423 2K 395K 2K

View Details

Co11ateral @co11ateral

6 days ago

Chinese Backdoor for Telecom Systems A while ago we showed the backdoor that the Chinese have been using to maintain persistence across telecom systems. The backdoor attaches itself to a raw network socket and inspects incoming traffic. It sees packets before firewall rules have a chance to process them. So even if your firewall is configured correctly, the backdoor can still see traffic that should have been blocked. To access the system hackers send a magic byte to get a reverse shell We also showed how you can detect it hackers-arise.com/compromising-t… @three_cube @_aircorridor

12 168 961 85K 758

View Details

forefy @forefy

a week ago

(Redteam) This is how macOS devs get compromised by opening a .docx (file opening a resume, opening a p.o. sow whatever) In short: > docx can have a remote template (feature to let it pull styles from an http server) > docx remote template runs VBA in your machine > this VBA can run an osascript like the one below >Word has sandboxing in place, but not if the file name starts with `~$.myfile` - they regexed it out for allowing Word itself to write temp files > macOS on start unzips any zip files /Users/youruser/Library > if that zip has a folder LaunchAgents with a plist file inside - that plist can invoke any command as a perfectly root level persistency (e.g. connect to a C2) I made this one open source as well, to increase awareness and improve the redteam around macOS endpoints to make sure EDR products raise the standards (and not, for example, trust the Word signed binary to write to login items as it pleases) Please share to increase awareness, and think twice when opening a Word file!

11 87 482 31K 454

View Details

hacksys @flopyash

2 weeks ago

[AI Security] Been deep into research on AI API abuse, and I found something interesting: LLM APIs can be repurposed as covert communication channels: dead drops, command execution, data exfiltration all blending into legitimate enterprise traffic. "RedBeacon AI"

1 1 10 1K 7

View Details

hacksys @flopyash

2 weeks ago

@cyberwarfarelab github.com/RedTeamOperati…

0 0 1 35 0

View Details

hacksys @flopyash

2 weeks ago

@RussianPanda9xx @cyberwarfarelab

0 0 0 14 0

View Details

hacker.house @hackerfantastic

3 weeks ago

How to find 0day using local LLM's and "recursive prompting". I used this approach on a NVIDIA DGX Spark using Qwen 3.6 heretic and LLAMA.cpp with LiteLLM and Openclaw. It successfully analyzed a complex FOSS code base and found a single high-risk RCE 0day.

15 150 864 77K 1K

View Details

Adam Chester 🏴‍☠️ @_xpn_

4 weeks ago

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! specterops.io/blog/2026/06/2…

16 187 621 109K 467

View Details

sapir federovsky @sapirxfed

4 weeks ago

Finally found the time to read this great blog post by @Thomasbyrne__ ! Its REALLY good, it describes NAA & FOCI, and a new CAP bypass. And the most important thing - the research process 😀 netspi.com/blog/technical…

1 16 55 4K 37

View Details

syndrowm @syndrowm

a month ago

Part one is here: blog.offensivecontext.com/abusing-ebpf-p…

0 1 2 151 2

View Details

Jeffrey Scholz @Jeyffre

a month ago

1 - So GLM 5.2 is 700b parameters (ish) 2 - 4x DGX Sparks can supposedly handle up to 700b parameters (give or take) 3 - GLM 5.2 is supposedly in striking distance of the performance of GPT 5.5 and Opus 4.8. In my brief tests, it's really not shabby at all. 4 - So for $20k, you can get near the frontier on your table. 5 - Extrapolate the trend, and you could have mythos/5.5 pro - class models in your dining room for the cost of a cheap car less than five years from now. Even without extrapolation, we're already the near frontier running locally. 6 - Paying real api costs, I could easily blow through $3,000 per month coding and running agents. The machine pays for itself in 6-7 months conservatively. 7 - In 3-5 years, most power users of AI will self-host. 8 - Am I missing something?

280 107 2K 523K 636

View Details

Mike Felch (Stay Ready) @ustayready

a month ago

As yall may have realized, I disappeared from the community for a little while we fight the most difficult fight of our life. My wife Angela was diagnosed with stage 3 cancer. We need all the help we can get, please consider supporting our fight. givesendgo.com/anchors-for-an…

10 39 113 22K 11

View Details

NVIDIA AI @NVIDIAAI

2 months ago

(2x DGX Sparks) + MiniMax M2.7 NVFP4 = 16 local AI agents running simultaneously 👀

mr-r0b0t @mr_r0b0t

2 months ago

16 local AI agents streaming at once! MiniMax M2.7 NVFP4 — 2x GB10, no cloud APIs.

61 30 368 181K 152

81 104 1K 142K 436

View Details

hacksys @flopyash

2 months ago

@cneuralnetwork Come to @cyberwarfarelab office at 19th main HSR

0 0 2 470 0

View Details

Cloudflare @Cloudflare

2 months ago

Cloudflare's security team spent the last few weeks testing Anthropic's Mythos against fifty of our own repositories. What we learned about offensive AI, why faster patching is the wrong reaction, and what the architecture around vulnerabilities has to look like next. cfl.re/49BRUqW

88 705 4K 1.7M 5K

View Details

𝕡𝕨𝕟𝕚𝕖 @0xpwnie

2 months ago

Just dropped my full notes on Pwn2Own Berlin 2026. Broke down the big wins by DEVCORE, the actual techniques they used, why these matter in the real world, and exactly where you can practice the same skills yourself. Full article here #Pwn2Own #P2OBerlin #CyberSecurity