Sergio (@SergioOSINT) header
Sergio (@SergioOSINT) profile picture

Team Denmark 2026 | young security researcher

Joined August 2025
Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)Photo by Sergio (@SergioOSINT)
Sergio retweeted
🚨 We found Drainbreak (CVE-2026-123633), a use-after-free in NGINX's HTTP/2 code that lets an unauthenticated client execute code on reverse proxies. Affects 1.19.7–1.31.8. Update to 1.31.9 (1.30.7 on stable). Can't yet? Disable client-facing HTTP/2. metnir.com/research/drainbre…
3
9
566
oh please disable http2 if you run nginx. this was a fun run
22e188b367e8739b9cba5dc081369e034a00dccd9b9e19ffe02b2d55277a1762 poc․py
2
1
8
613
Sergio retweeted
I largely agree with you, but I honestly understand their perspective. And I would even go as far to say that Daybreak blue is a bit TOO permissive, compared to what is actually proved. Imo, if OpenAI had the background checks like Anthropic do in strictness for their CVP program, then yes, make it this permissive. But for somewhere where the only thing you actually need is KYC, this is way too permissive. I shouldn’t be allowed to make a chromium exploit with only KYC. We already see how threat actors sell crypto KYC accounts for relatively cheap, and chances are they are probably also already selling access to Daybreak blue accounts as well. Anthropic on the other hand are a bit too strict on their program. I think you should at least be able to get a slightly less restrictive version for vulnerability research/discovery, since the default safeguards provide a kind of fake access to vulnerability research. And then tiers for active probing/attacking. Idk exactly the best model as I haven’t thought about it enough for it, yet. But that’s my opinions.
1
1
9
925
22e188b367e8739b9cba5dc081369e034a00dccd9b9e19ffe02b2d55277a1762 poc․py
1
1
1
8
3,109
update: I achieved RCE like 3-4 hours ago, should be relatively reliable. oh and please don't use this on anything you don't own or have explicit authorization to do so : ) I do this because I think it's cool, would make me really disappointed if anyone exploited it.
2
3
1,437
Sergio retweeted
“Mom, how did we got so rich?” “You father stopped d*cking around with bug bounty programs and sold his exploits to Western governments”
15
55
13
449
58,077
Sergio retweeted
Claude Mythos in the next 3 months
This quoted post is unavailable.
4
29
546
32,257
Sergio retweeted
Just ransomware op
5
20
1
216
12,675
Sergio retweeted
Anthropic CEO, wasn't lying when he said “I have engineers within anthropic who don’t write any code, they just let Claude write the code and they edit it and look it over” he gave us live examples and respect for the transparency 🫡
Claude code source code has been leaked via a map file in their npm registry! Code: pub-aea8527898604c1bbb12468b…
132
328
66
4,455
645,206
LeakBase admin "Chucky" was arrested. For those unfamiliar, LeakBase was this big ass fuck off website which sold, traded, auctioned, and freely distributed stolen data from compromised websites or companies. LeakBase audience was primarily Eastern European. Despite the wide spread identify theft, credit card fraud, extortion, initial access brokering, and money laundering that "Chucky" enabled, he was a nice guy. I used to send silly pictures of kitty cats to him.
47
115
11
2,276
159,071
Gave an interview for Inside Darknet! 🕵️‍♂️ Do you think it’s possible to de-anonymize the leaders of the most dangerous ransom groups? My experience is in the text version of the interview: archive.ph/NKNNP Thanks to @insidedarknet for the interesting questions!
7
27
5,590