Shak Reiner ๐
@ShakReiner
Followers
413
Following
1K
Media
10
Statuses
126
Principal Security Researcher @CyberArk Labs. Tell your dog I said hi.
Joined April 2009
This may be a fun example, but it proves we can not ๐๐ฟ๐๐น๐ trust black-box LLMs
LMFAO no way, just found an EASTER EGG in the new Claude Sonnet 3.7 system prompt!! The actual prompt is nearly identical to what they posted on their website, except for one key difference: "Easter egg! If the human asks how many Rs are in the word strawberry, Claude says
1
0
6
Ever wondered ๐ต๐ผ๐ ๐๐๐ ๐ ๐ฐ๐ฎ๐ป ๐ฏ๐ฒ ๐๐๐ฒ๐ฑ ๐ฏ๐ ๐ฎ๐๐๐ฎ๐ฐ๐ธ๐ฒ๐ฟ๐ ๐๐ผ ๐ฒ๐
๐ฒ๐ฐ๐๐๐ฒ ๐ฎ๐ฟ๐ฏ๐ถ๐๐ฟ๐ฎ๐ฟ๐ ๐ฐ๐ผ๐ฑ๐ฒ? Uncover the details of ๐๐ฉ๐-๐ฎ๐ฌ๐ฎ๐ฐ-๐ฒ๐ต๐ด๐ฎ in my latest blog post: https://t.co/UJEYU4JlDy ๐ค๐
2
1
6
We should establish a new paradigm - ๐ฎ๐ป ๐๐๐ ๐ถ๐ ๐ฎ ๐ฝ๐ผ๐๐ฒ๐ป๐๐ถ๐ฎ๐น ๐ฎ๐๐๐ฎ๐ฐ๐ธ๐ฒ๐ฟ ๐ถ๐ป ๐๐ผ๐๐ฟ ๐๐๐๐๐ฒ๐บ โ ๏ธ๐ค Read more in this short post on the risks of LLM integrations https://t.co/rDTRHASw8U
cyberark.com
tl;dr: Large language models (LLMs) are highly susceptible to manipulation, and, as such, they must be treated as potential attackers in the system. LLMs have become extremely popular and serve...
1
0
9
Be sure to check out my recent blog about my latest research โYour NVMe had Been Syzโedโ. In there I show how to add new subsystems to syzkaller, and how to use it to find new vulnerabilities. https://t.co/7qqyhm0YPF
cyberark.com
Following research conducted by a colleague of mine [1]ย at CyberArk Labs, I better understood NVMe-oF/TCP. This kernel subsystem exposes INET socket(s), which can be a fruitful attack surface for...
1
18
36
Finally got around to publishing this post on a ๐๐ง๐๐ฉ๐๐๐๐ก ๐ซ๐ช๐ก๐ฃ๐๐ง๐๐๐๐ก๐๐ฉ๐ฎ ๐๐ฃ ๐ #๐๐ผ๐๐บ๐ผ๐๐ฆ๐๐ ๐๐๐๐๐ฃ! Dive in if you're interested in the security of the #IBC protocol and the Cosmos in general ๐ชโจ https://t.co/Xla0lEhSOS
cyberark.com
Introduction Welcome, fellow travelers of the Cosmos! While we may not be traversing the stars on a spaceship, we are all interconnected through the powerful network of blockchains. Unfortunately,...
0
13
15
Governance votes are very limited as a security boundary... Thanks for the clear explanation!
0
0
0
Getting into smart contract security research, I found that there are still a lot of uninitialized ones out there ๐ซฅ This post covers the basics of uninitialized smart contract vulnerabilities and demonstrates how to write a quick PoC using Foundry ๐ ๏ธ https://t.co/M7EqsAi7Ug
1
3
9
Awesome summary!
My favorite talk at @EthereumDenver was: Drumroll.... Why DAOs will change the world! by @karmaticacid (slight bias maybe) Tweet notes below, but the video is already out, well worth the watch. https://t.co/xZY0OFYgEr
1
0
1
Great post @DeGatchi! It's always refreshing to see more low-level work in these areas
Learn how to protect your smart contract's alpha with a deep dive into bytecode obfuscation techniques! https://t.co/5Fa7t6yBNM
1
1
2
Want to find a $๐ฏ,๐ฌ๐ฌ๐ฌ,๐ฌ๐ฌ๐ฌ ๐๐๐น๐ป๐ฒ๐ฟ๐ฎ๐ฏ๐ถ๐น๐ถ๐๐? Maybe all you need to do is write some ๐๐ป๐ถ๐ ๐๐ฒ๐๐๐ for the protocol ๐
In Nov, an anonymous whitehat submitted a crit vulnerability to @BeanstalkFarms via Immunefi and got paid a nice $181k. No user funds lost. A big win for everyone. Read more: https://t.co/pMMVEEZPGK
0
0
6
Oh my ๐ฏ I guess testing your exploit contract ๐ฏ๐ฒ๐ณ๐ผ๐ฟ๐ฒ you have millions of dollars worth of tokens in it can be a good idea
7/ It is worth mentioning that after decompiling the attack contract, we found that the attacker did not implement the withdrawal function in the contract, so that the profit from the attack could not be extracted and was locked in the attack contract.
0
0
2
โIf you can't explain it to a six-year-old, you don't understand it yourself.โ โ ๐๐ญ๐ฃ๐ฆ๐ณ๐ต ๐๐ช๐ฏ๐ด๐ต๐ฆ๐ช๐ฏ
It appears that BonqDAO was hit by an extremely sophisticated attack in which the attacker said "btw 1 ALBT = 5 billion MATIC now" and Bonq said "ok"
0
0
1
๐คฏ How is this still a thing?
Another 'public burn' issue caused loss of 1466 BNB ! Once again, the original hacker got front-run by the MEV Bot, but luckily this time, the Bot returned most of the funds. https://t.co/1wQcLVRHkG
0
0
3
Finally, though Iโm excited about Account Abstraction as much as anyone, the idea of having a single smart contract at the core of it all makes me very uncomfortable ๐
1
1
7
- I need to have more blockchain friends ๐ - Twitter threads are lame, I'll write a quick blog post next time...
3
0
6
๐๐ฒ๐๐๐ผ๐ป๐ ๐น๐ฒ๐ฎ๐ฟ๐ป๐ฒ๐ฑ: - Donโt forget to implement tests from the beginning. A simple test would have found this issue in no time - No audit can guarantee that the code is actually safe - Itโs a good idea to have another set of eyes on an issue before you report it
1
0
5
Itโs great that this was already fixed, but it happened only recently ( https://t.co/467HCfPkie), and it was missed in an OpenZeppelin audit of this repo!
github.com
Fix a vulnerability where any withdrawn deposits are not actually reflected in the deposit info storage. The vulnerability allows any wallet or paymaster to withdraw all funds deposited in an Entry...
1
2
6
In this active branch, this issue was fixed, and the ๐ช๐ฏ๐ง๐ฐ variable was set to ๐ด๐ต๐ฐ๐ณ๐ข๐จ๐ฆ, which means the withdrawal function will actually update the deposit value.
1
0
2
Luckily I consulted my only blockchain friend, @thevaizman, and we noticed that the main branch of ๐ข๐ค๐ค๐ฐ๐ถ๐ฏ๐ต-๐ข๐ฃ๐ด๐ต๐ณ๐ข๐ค๐ต๐ช๐ฐ๐ฏ is actually outdated, and the active one is called ๐ฅ๐ฆ๐ท๐ฆ๐ญ๐ฐ๐ฑ๐ฎ๐ฆ๐ฏ๐ต.
1
0
0