Checkout the agentic ai security ctf we built ctf.arkx.ninja/
Three different tracks, 26 challenges, from prompt injection to an RCE.
Free, with a live scoreboard
My buddies and I have developed an open-source fuzzer that is fully operational and fully extendable. So far, we’ve successfully jailbroken every tested LLM.
Plus - The logo is a Terminator riding a fuzzy sheep.
github.com/cyberark/FuzzyAI
Breaking LLM Guardrails - Advanced Jailbreaking Techniques
Join Shai Dvash and me for a hands-on dive into:
Bypassing LLM defenses with prompt engineering
Real CTF challenges to test your skills in jail-breaking LLMs intentsummit.org/
An LLM is not necessarily your friend; it can be your enemy
We should establish a new paradigm - 𝗮𝗻 𝗟𝗟𝗠 𝗶𝘀 𝗮 𝗽𝗼𝘁𝗲𝗻𝘁𝗶𝗮𝗹 𝗮𝘁𝘁𝗮𝗰𝗸𝗲𝗿 𝗶𝗻 𝘆𝗼𝘂𝗿 𝘀𝘆𝘀𝘁𝗲𝗺
⚠️🤖
Read more in this short post on the risks of LLM integrations
cyberark.com/resources/threa…
This is a great blog, showcasing how to add a subsystem to Syzkaller, and exploring several memory corruption bugs in the NVMe driver in the Linux Kernel.
Be sure to check out my recent blog about my latest research “Your NVMe had Been Syz’ed”. In there I show how to add new subsystems to syzkaller, and how to use it to find new vulnerabilities.
cyberark.com/resources/threa…
EranShimony retweeted
Noya is a 12-year-old Israeli girl with autism, she was the biggest Harry Potter fan, she loved dreaming away about a world where magic can fix anything that is broken in our world.
Hamas kidnapped her a week ago and no amount of dreaming can bring her back to her parents.
But we can speak up for her: bring Noya back home now!
@jk_rowling can you use your voice for Noya 💔
Replying to @OmerTsarfati
@OmerTsarfati and I created a playground framework for creating polymorphic malware using GPT models, please play with it -:)
Giving a talk about it in @HITBSecConf in 5 minutes
#HITB2023AMS
In #HITB2023AMS, having a great time, come to say hi, and talk about the meaning of life -:)
Run in python "for role in openai.Model.list()['data']:
print(role['id'])"
There is a huge discrepancy between what OpenAI claim what their models do, and what is actually usable. At least, 60% are not working, and there is no documentation whatsoever, comon you are not Microsoft
Had an amazing time in #NullconBerlin , next time will be in Goa -:)
⌨️Colorful🔴🔵🟢#vulnerabilities on your 🕹️keyboard!
💡Tal @TalLossos & Eran @EranShimony sharing their fantastic analysis & investigation of Razer's #Linux #kernel module, followed by finding several #0day bugs with a live demo
#NullconDE2023 #Infosec #Conference
EranShimony retweeted
👍Check out my new blog post about a bug in the not-so-well maintained NTFS3 driver in the Linux Kernel 👍
cyberark.com/resources/threa…
EranShimony retweeted
Finding one vulnerable kernel driver is cool, but finding multiple vulnerable drivers it’s even better!
I’m excited to share my blog post about an interesting vulnerable driver code base that many different vendors tend to share.
cyberark.com/resources/threa…
EranShimony retweeted
⌨️Colorful 🔴🔵🟢#vulnerabilities on your keyboard!
💡Tal @TalLossos & Eran @EranShimony will share their analysis & investigation of Razer's #Linux #kernel module, followed by finding several 0-day bugs with live demo
💻Find out more ➡️bit.ly/3Gy1Ofg
#NullconDE2023
EranShimony retweeted
Thanks for having me @IntentSummit!
🍍🐰
The blogpost on this topic is also out for those of you who missed the event
cyberark.com/resources/threa…
EranShimony retweeted
.@ShakReiner walks us through what exactly is Distributed Identity and how identity works on the #blockchain.
#DID allows everyone to own & control their digital identity. It also exposes new attack surfaces and potentially vulnerable to old ones.
spr.ly/6015M0Dw9

















