@gN3mes1s Profile picture

Giuseppe `N3mes1s`

@gN3mes1s

windows, macos, linux, android && lowlevel && ring-1 lover; EDR chef; malware hunter; purple team💜

Similar User
Seongsu Park photo

@unpacker

Bart photo

@bartblaze

Joshua J. Drake photo

@jduck

Xylitol photo

@Xylit0l

REhints photo

@REhints

James Forshaw photo

@tiraniddo

Didier Stevens photo

@DidierStevens

`Ivan photo

@Ivanlef0u

Evilcry_ photo

@Blackmond_

volatility photo

@volatility

Paul Rascagnères photo

@r00tbsd

Alex Ionescu photo

@aionescu

Matt Nelson photo

@enigma0x3

Jonathan Salwan photo

@JonathanSalwan

Ange photo

@angealbertini

Giuseppe `N3mes1s` Reposted

Following the release of our latest system level safeguards, today we're sharing new research papers outlining work on Llama Guard 3 1B & Llama Guard 3 Vision that support input/output safety on the edge and in multimodal prompts.

Tweet Image 1

Giuseppe `N3mes1s` Reposted

🚨 We discovered two malicious Python packages in #PyPI repository that remained undetected for over a year. These packages mimicked tools for working with popular AI language models (#ChatGPT and #Claude), silently exfiltrating data and compromising developer environments.…

Tweet Image 1

Giuseppe `N3mes1s` Reposted

Just posted: "Empowering Gemini for Malware Analysis with Code Interpreter & GTI". In our 3rd installment, Gemini successfully deobfuscated IOCs and enriched analysis with GTI function calls, avoiding hallucinations. More advancements coming soon! cloud.google.com/blog/topics/th…


Giuseppe `N3mes1s` Reposted

🎉 It's official! The #CAMLIS2024 abstracts & videos are now live on our website & our YouTube channel! 🚀 Don't miss this opportunity to explore some incredible work at the intersection of machine learning & info security! 🌐 Website: camlis.org/2024-conference

Tweet Image 1
Tweet Image 2
Tweet Image 3

Giuseppe `N3mes1s` Reposted

Quick demo showing how, with the latest macOS ChatGPT app, a prompt injection can now come from random text in the Terminal. In this case the prompt injection invokes the memory tool (to=bio) to persist information into your ChatGPT's brain.


I mean. What could go wrong.

ChatGPT 🤝 VS Code, Xcode, Terminal, iTerm2 ChatGPT for macOS can now work with apps on your desktop. In this early beta for Plus and Team users, you can let ChatGPT look at coding apps to provide better answers.



Giuseppe `N3mes1s` Reposted

Backdoor attempt on @exolabs through an innocent looking PR. Read every line of code. Stay safu.

Tweet Image 1

Giuseppe `N3mes1s` Reposted

Getting made fun of because you cover laptop webcam with a sticker? 😭 Here are materials from my talk about controlling ThinkPad X230 webcam LED over USB presented at POC by @POC_Crew 😎 Use these as a comeback 😁 Slides: docs.google.com/presentation/d… Code: github.com/xairy/lights-o…

Tweet Image 1
Tweet Image 2
Tweet Image 3
Tweet Image 4

Giuseppe `N3mes1s` Reposted

You either detect the procedure or you don’t. Does it matter what ID it’s mapped to?

Hear me out… does anyone else think the MITRE ATT&CK framework is basically a waste of time? Feels like people care more about ‘Did you map it to MITRE?’ than actually understanding what’s happening. Are we here to do real security work or just fill out checklists?



Giuseppe `N3mes1s` Reposted

#flareon11 For task 9, massive thanks to the hero who wrote Python bindings for the Time-Travel Debugger (github.com/commial/ttd-bi…). Being able to record a process execution once and then arbitrarily navigating it (reading regs/memory along the way) really feels like a superpower.


Giuseppe `N3mes1s` Reposted

🚨EDR telemetry evals for the new Linux category have started. The first results are in, and they ain't pretty 😔 I genuinely hope vendors succeed because it benefits everyone! Even so, the telemetry evals will remain objective and open to improvement via community feedback. We…

Tweet Image 1

Giuseppe `N3mes1s` Reposted

"Anton’s Alert Fatigue: The Study" bit.ly/3UJWBI1 is a result of some agonizing pondering of the "alert fatigue problem" over the years. Hopefully it is a fun read, because it was not a very fun write -)


Giuseppe `N3mes1s` Reposted

LLMs as vulnerability hunters? Yup. Our Vulnhuntr tool from @ProtectAICorp uses Claude to scan Python code for 0days. 🤯 Check out the NoHat talk for the full story from @DanHMcInerney & @byt3bl33d3r 👇 hubs.ly/Q02Xqxkv0


Giuseppe `N3mes1s` Reposted

Test-driven development—not just for software engineering. @matterpreter breaks down how applying this logic streamlines how you execute, evaluate, and iterate on your detections to better augment your defensive coverage hubs.la/Q02W9xs80

Tweet Image 1

Giuseppe `N3mes1s` Reposted

We've observed a rise in Living off the Land email attacks where attackers abuse legitimate service infrastructure. Our newest Attack Spotlight details one of these attack variants abusing Docusign to deliver malware via callback phishing: sublime.security/blog/living-of…

Tweet Image 1

Giuseppe `N3mes1s` Reposted

I'm delighted to share with you a #Rust project we've been working on for a while called #Hyperlight. The Hyperlight project enables devs to use KVM or Hyper-V as a library for running hypervisor isolated funcs with extremely low-latency cold starts. opensource.microsoft.com/blog/2024/11/0…


Giuseppe `N3mes1s` Reposted

I'm hiring for a Head of AI Security Testing (aka AI red teaming) role! This role is responsible for conducting applied research to develop novel ways to manipulate and breach AI systems, focused on meaningful impacts to safety and security. Joining a huge team of hackers,…


The 6.3 Risks and Mitigations part of the paper are so eye opening from a Risk & Security POV. "it is imperative that agents operate under a strict principle of least privilege, and maximum oversight." Just to quote some of the important points we are going to see more and…

📢Introducing Magentic-One, a generalist 5-agent multi-agent system for solving open-ended web- and file-based tasks. 🤖🤖🤖🤖🤖 Magentic-One represents a significant step towards agents that can complete tasks that people encounter in their daily lives and can achieve strong…

Tweet Image 1


Giuseppe `N3mes1s` Reposted

🍅Surprising finding: Basic adversarial pop-ups trick state-of-the-art VLMs (e.g., @AnthropicAI computer use agent) into clicking 🚩>90%🚩of the time in OSworld! 🥝Clear signal: We need more robust safety measures before deploying computer use agents at scale.

Tweet Image 1

Humans sometimes get distracted by pop-ups… but for AI agents, it’s worse! Pop-ups explicitly designed for agents can make them click 87% of the time, majorly derailing their tasks. @taoyds @Diyi_Yang arxiv.org/abs/2411.02391 github.com/SALT-NLP/Popup…

Tweet Image 1


InjecGuard: Benchmarking and Mitigating Over-defense in Prompt Injection Guardrail Models injecguard.github.io


Loading...

Something went wrong.


Something went wrong.