honu

memento vivere

[Johann Rehberger @wunderwuzzi23]

-> deterministic software (a machine)
-> agentic software (an actor)
-> organic software (an organism)

deterministic software can be exploited
agentic software can be hijacked
organic software can be infected

future will be wild, the attack surface is alive

[honu @honuonhval]

@ThePrimeagen Does this work for typescript?

[honu @honuonhval]

@vxunderground @nikhil_mitt @struppigel CARTP not cardi-b please

[honu @honuonhval]

@vxunderground @TorGuard Very valuable resource and very generous. Thank you

[honu @honuonhval]

Oh! It’s been a year since my first flare-on challenge. Let’s see how much I have improved since then

[Mandiant (part of Google Cloud) @Mandiant]

The 12th Annual Flare-On Challenge kicks off Sept 26 at 8PM EST!

Reverse engineering pros, from Windows to Web3 (with a YARA twist), it's your time to shine. 🏆

Get ready → https://t.co/8O6q9qhgvl

#FlareOn12

[honu @honuonhval]

@arcanuminfosec Would love to be picked. I've been doing my own studies but would love to learn from one of the best in the industry.

[honu @honuonhval]

Coding agents are vibe bots. Having said that I’m in the camp of “let’s figure out how to control it to build what I want”

[JS0N Haddix @Jhaddix]

⚠️ Giveaway time! ⚠️ 👇

📢 Our new course "Attacking AI" will be Feb 27-28!

This two-day course equips security professionals with the tools and methodologies to identify vulnerabilities in AI systems. It's gonna be a BANGER.

Syllabus: https://t.co/cY9vcI7Z5y

We are giving away two seats this week!

⁉️How to enter the giveaway:

♻️ Repost this post = 2 Entries
🗣️ Reply = 1 Entry
❤️ Like = 1 Entry

[honu @honuonhval]

@OpenAI 🔥🔥🔥 and to the stars!!!

[honu @honuonhval]

@vxunderground @whid_ninja Love to get my hands wet with hardware hacking

[honu @honuonhval]

@vxunderground @_MG_ @Hak5 Oh! Last minute love! My favorite

[honu @honuonhval]

@vxunderground Twisted Metal was such a jaw dropping game when it first came out. That was pure joy

[honu @honuonhval]

@vxunderground Yes sir. That’ll do

[honu @honuonhval]

@vxunderground @MalDevAcademy “Would you like to know more?”
Yes, yes I do!

[honu @honuonhval]

@vxunderground @0verfl0w_ Yes I love studying all things malware!

[honu @honuonhval]

This is quite interesting. I’ve always enjoyed achieving my goals by first defining my goal then working backwards figuring out what conditions are needed for that goal. Defining sub-goals and smaller tasks needed for that and so on

[elvis @omarsar0]

Reverse Thinking Makes LLMs Stronger Reasoners

Shows that training LLMs to learn "reverse thinking" helps to improve performance in commonsense, math, and logical reasoning tasks.

It claims to outperform a standard fine-tuning method trained on 10x more forward reasoning.

[JS0N Haddix @Jhaddix]

🛑 GIVEAWAY ALERT 🛑⬇️

Today is day SEVEN of EIGHT days of @arcanuminfosec and friends Black Friday and Cyber Monday giveaways!

Today @DanielMiessler and I are giving away one seat to his next "AUGMENTED" Course!

Dan runs a FANTASTIC course on integrating AI into the very fabric of your life, future-proofing yourself, and maximizing your potential using AI.

See: https://t.co/OlTLUEb6u1

Next run should be around Feb!

You know the drill! Each person can have up to FOUR entries to the giveaway!

♻️ Share This Post = 2 Entries
♥️ Like This Post = 1 Entry
✍️ Comment This Post = 1 Entry

Winners will be posted next Tuesday!

[honu @honuonhval]

@daeken Would love to learn more about your approach if you are ok with sharing! Just getting into vuln research so I had to even look up what you meant by sources and sinks in this context :) Time for me to dig in deeper

[radare @radareorg]

And this is how #frida and #radare2 look together on a browser, so as Ole says, every Frida user can become a radare user over the night
#r2con2024

[honu @honuonhval]

Been thinking about creating tool for people to create teams agents optimized for different workflows. This looks like something I need to study

[elvis @omarsar0]

AFlow: Automating Agentic Workflow Generation

AFlow is a novel framework for automating the generation of agentic workflows.

It reformulates workflow optimization as a search problem over code-represented workflows, where LLM-invoking nodes are connected by edges.

AFlow efficiently explores the search space using a variant of MCTS, iteratively refining workflows through code modification, tree-structured experience, and execution feedback. It introduces operators that encapsulate common agentic operations (like Ensemble, Review & Revise) to enhance search efficiency.

Experiments across six benchmark datasets demonstrate AFlow’s effectiveness, showing a 5.7% improvement over manually designed methods and a 19.5% improvement over existing automated approaches. AFlow also enables smaller models to outperform GPT-4o on specific tasks at just 4.55% of its inference cost.

The framework maintains strong performance even without predefined operators, demonstrating its ability to discover effective workflow structures.

It's a compelling approach especially because it seems to work on different kinds of tasks and can potentially discover more optimal ways to optimize costs for agentic workflows. Not sure about latency but it's also interesting that they can efficiently get smaller models to outperform larger ones.