I am in favor of strong regulation for AI but ONLY if we get to call them the Turing Police

just by turning off the 'code error' feature in claude's neural net, it will automatically start... fixing errors

that's the power of mechanistic interpretability. imagine what other capabilities are hidden in the weights language models

This is a fantastic meta-paper if you're writing a paper on a novel fuzzing technique: mschloegel.me/paper/schloege…

In the fuzzing session, Moritz Schloegel is presenting our paper on potential pitfalls of evaluating fuzzers. We checked 150 papers published in the past six years and performed 8 case studies to show what can go wrong during an evaluation. We publish all data related to our ...

Directing fuzzers to well-defined locations is an important topic for patch verification, 1-day PoC creation, and debugging. Instead of 'just' adjusting scheduling, we argue that applying constraints during input generation further improves the results. @IEEESP

Today I’ll be at IEEE Security and Privacy IEEE S&P with Andrew Carney from ARPA-H to discuss AIxCC. There’s so much fantastic research being presented on how AI can be used for security, we’re excited to join the fun 🎉

🚀 Introducing Cover-Agent 🧪

An open-source tool that includes a reimplementation of Meta's TestGen-LLM for automatically enhancing test suites.

Manager: 'We must improve old test suites for better code coverage. Can you handle it?'
Me: 'Sure, my favorite task... (Not!) 🤷‍♂️'

I can't say I remember consenting to this study

I would be more impressed if they were sending mail from the present to the past but okay

Our systemization of knowledge paper on „Prudent Evaluation Practices for Fuzzing“ (arxiv.org/abs/2405.10220) has received a Distinguished Paper award at IEEE #SP24 🎉

'Can LLMs be used as helpful security assistants for vulnerability detection?'

Our tool 'SecLLMHolmes', published at IEEE S&P '24, answers this question by evaluating capabilities of LLMs for identifying and reasoning about security vulnerabilities.

computer.org/csdl/proceedin…

I released rr 5.8.0. Lots of bug fixes as usual, plus initial support for LLDB. github.com/rr-debugger/rr…

Today our book with Farenain turns 1 year old! Keep Fuzzing Against the Machine.

Many thanks to all our supporters. And a gigantic community.

nSinus-R (@[email protected]) Nikias Bassen Aurélien Francillon Dave Aitel Davide Balzarotti Trung Nguyễn Hoàng Halvar Flake lcamtuf (@[email protected]) Brendan Dolan-Gavitt Advanced Fuzzing League
Without you this could't happen!

Operation ParkHack begins

Yes. It's never been a better time to be the kind of person who's obsessive about wording and editing details! Unleash that part of you that loves using a red pen 😈

LLM prompt engineering is getting really sophisticated, wow