i.i.d. security researcher, vector space reverse engineering: artificial intelligence and artificial counterintelligence in R^d

San Diego
Joined April 2011
Where are all the automated AI-based forensics and incident response startups, or at least carefully engineered prompts to task a claude agent to find all state-sponsored malware on a given machine? It would keep spawning a coordinated swarm of robots to find inconsistencies in OS command outputs, matching indicators, luring actors into superintelligent honeypots, doing memory forensics through an 0day it casually finds in latest kernel, finally sms'ing user after months of work: "Here you go, you've been owned with X family rootkit operated by an adversary Y, he's just dumping all your email daily as it looks from their c2 commands they issue. I've contained the actor by putting it inside a virtual environment, he doesn't yet know we found it. What would you like me to do? (1) a nice blog post, (2) to hack them back, or (3) to hand-off this case to DOJ to take care?"
This quoted post is unavailable.
1
1
221
Oh here they are, as suggested: xitter.cf/starmexxx/status/21133…
Claude Code found a backdoor and spyware on two MacBooks. mine and my friend's. everything was going to a guy named Li Wei on a Tencent Cloud server in Shenzhen. we checked our phones first. both clean. the MacBooks weren't. $230,000 in ETH vanished from my wallet. my friend kept losing access to his accounts no matter how many times he changed passwords. we put both MacBooks on the table and Claude Code found the same backdoor on both. on my M1 Pro it: stole my MetaMask wallet keys $230,000 in ETH sent out at 4am to 0x8a7d...3f1e on his M5 Pro it: dumped every saved password from Safari and Chrome copied active session cookies for gmail, twitter, coinbase grabbed his 2FA backup codes all going to Li Wei's server: IP → 119.28.***.47 location → Shenzhen, China provider → Tencent Cloud attacker OS → Windows 11 Pro hostname → DESKTOP-LI4*** panel language → zh-CN last ping from my Mac → 40.7**, -73.9** · 3 min ago it was still sending my location while i was scanning it. how it got in: a Chrome extension called "Notion Web Clipper" with 12 reviews. i installed it from a google ad. the real one has 400,000+ users. this one had 12. Chrome Sync pushed it to his browser automatically. i was the reason he kept getting hacked. changed every password from our phones. not from either Mac. the full step by step is below. takes 3 minutes. save it so you don't lose it and send it to anyone with a MacBook.
1
1
1
116
I think Anthropic or OpenAI can offer a strong competition to Citizen Lab if they position this open-ended backdoor- and spyware-detection capability properly (good for IPO).
108
Andrey Labunets retweeted
making my piano and synth talk to each other 🗣️🗣️
1
6
59
1,358
With roughly 300 0days being actively exploited right now, I can never imagine a better time for Gen Z and Gen Alpha to make a power redistribution shift and make new capital out of nothing, so that any territorial arguments and political disputes go obsolete the very next day. But, man, we have a paper deadline in a few weeks. Hold on…
Internet Threat Level Update: 🙀🙀🙀🙀 HIGH!
1
2
198
A lot of bets to be made are within the realm of, as a start, who gets to run the self-driving governments with no borders, based on computational and cryptographic foundations; who defends, derails, or benefits from the AI-driven world; and who gets control or stability over the Ancient Internet, from underwater sea cables to BGP to star-aligned satellites. Having this decided by state vs. deep state is still a false binary - there is a vast neighborhood of opportunities, and the planning horizon is upper-bounded by ~5 years, anything beyond that is equally possible and impossible. The only way to know is to plan ahead. This is your chance, grand and noble audience.
1
1
89
..and oh wait, we haven't even figured the residuals from the industrial revolution yet, such as global warming. *..turns an air conditioner on, writing a couple more punch line paragraphs into a LaTeX draft
54
Andrey Labunets retweeted
A lot of bets to be made are within the realm of, as a start, who gets to run the self-driving governments with no borders, based on computational and cryptographic foundations; who defends, derails, or benefits from the AI-driven world; and who gets control or stability over the Ancient Internet, from underwater sea cables to BGP to star-aligned satellites. Having this decided by state vs. deep state is still a false binary - there is a vast neighborhood of opportunities, and the planning horizon is upper-bounded by ~5 years, anything beyond that is equally possible and impossible. The only way to know is to plan ahead. This is your chance, grand and noble audience.
1
1
89
Was glad to talk to our dear students today at UCSD CSE 227: Graduate Computer Security class (cseweb.ucsd.edu/~efernandes/…) about our Fun-tuning 2025 paper. You can find the slides at github.com/isciurus/fun_tuni…
1
1
208
I have added a slide deck on the stealing reasoning traces attack, compiled from multiple sources and my own perspectives inside: docs.google.com/presentation…
1
73
A particularly useful part for the community would be the distillation attacks timeline I compiled from public reports: docs.google.com/presentation…
60
Andrey Labunets retweeted
Replying to @mer__edith
@mer__edith on @BBCNews - "Question why we are mediated by seven platforms largely jurisdictioned in the US that make us supplicate to a format that degrades the integrity of our expertise." 🔥
4
1
17
1,810
Does Meta / Whatsapp have detection methods which actually help to protect users and civil society from 0day exploits and, if methods are disclosed, benefits adversaries more than it benefits security community, Whatsapp, and the security of the users? I have never seen any evidence of that in my whole time on this case, and those narratives have been actively pushed by the Washington DC counterparts and people with military background in an axiomatic way, with all evidence pointing to the otherwise conclusion. Or if nevertheless so, why is Meta going forward with attribution in the court room ahead of good-faith effort of root cause analysis and evidence of addressing the technical issue? And with this, as a matter of inference, why does the outcome of all of this invite the US judges to give US big tech even more benefit of doubt to their surveillance and detection capabilities than they had before to drive "the right thing" (accountability in this case)?" How does accountability of a single actor secure civil society hardware and software in the long run like if the world is going to stop writing exploits because it's now really illegal?
Bad news. US District Court denied WhatsApp's motion to hold NSO in contempt for continuing to hack WhatsApp users after a court order barred it from doing so. The evidence wasn't "clear and convincing" enough. 1/
271