samet g. retweeted
*sshole is out
if you run sshd on alpine, openwrt or any musl box with password auth on: update now
official writeup: sshhole.com
the model isn’t the moat. the harness is. my recon, my old reports, attack surface nobody else is mapping, all fed in so it sees what others miss. welcome #yeswehack and #bugbounty
samet g. retweeted
Şahsen bunun doğru olduğuna inanmıyorum, claude arada hem nalına hem mıhına vurmuş.
You open DeepSeek.
You type a prompt.
You think DeepSeek is answering you.
It wasn’t.
Your conversation was sent to Claude through thousands of fake accounts.
They claim showed you Claude’s answer and kept the data, That’s just one finding from Anthropic’s new threat report.
The rest is even more unsettling.
samet g. retweeted
Everyone benchmarks CTI models the same way: a multiple-choice exam. GPT-5.4 gets 82%. Gemini 3.5 Flash gets 81.6%. Impressive, right up until you remember that no analyst's actual job is answering four-option questions about ATT&CK.
So we handed those same models a real threat report and asked which techniques were in it. GPT-5.4 fell to 46.5. Gemini to 28.2. The quiz was never the job.
We built BeyondCTI-27B for the job. On that same extraction task it scores 62.3, nearly 16 points ahead of GPT-5.4 and more than double Gemini, at 27B, open weights. It already runs inside our own pipeline, reading forums and leak sites at 4am so an analyst doesn't have to.
Apache 2.0, on Hugging Face. Run it on your own reports and tell us where it breaks. Full writeup below.
Çoğu CTI ürününde "yapay zeka" dedikleri şey aslında birinin API'si. Biz onu yapmak istemedik.
Kendi modelimizi eğittik, kendi verimizle besledik, kendi hattımızda çalışıyor. Gerçek analiz işinde frontier modelleri geçiyor, hem de Türkiye'den. Bu ölçekte iş burada da yapılıyormuş, kendi elimizle görmek güzel.
Everyone benchmarks CTI models the same way: a multiple-choice exam. GPT-5.4 gets 82%. Gemini 3.5 Flash gets 81.6%. Impressive, right up until you remember that no analyst's actual job is answering four-option questions about ATT&CK.
So we handed those same models a real threat report and asked which techniques were in it. GPT-5.4 fell to 46.5. Gemini to 28.2. The quiz was never the job.
We built BeyondCTI-27B for the job. On that same extraction task it scores 62.3, nearly 16 points ahead of GPT-5.4 and more than double Gemini, at 27B, open weights. It already runs inside our own pipeline, reading forums and leak sites at 4am so an analyst doesn't have to.
Apache 2.0, on Hugging Face. Run it on your own reports and tell us where it breaks. Full writeup below.
8
611
samet g. retweeted
CTI analysts, we see you. Twelve browser tabs, six forum logins, a Tor session that dies every twenty minutes, and a notes file called "final_final_v3.txt". That's the industry standard workflow for dark web monitoring in 2026, apparently.
So we built the boring solution: one stream. 100+ dark web forums, onion sites included, scraped continuously into a single platform. Roughly a million raw threads so far, growing every hour. No logins, no Tor babysitting, no tab archaeology. You search, you read, you get on with your day.
We're opening researcher access soon. Very few people collect at this scale, and fewer still let anyone else look at it. DMs are open.
Tek forumda bakarken sıradan görünen adam, dördüncü forumda başka isimle aynı postu atınca başka bir şeye dönüşüyor. Ben ilk seferinde iki ayrı kişi sandım, sonra desen tuttu.
Tek tek gezen analist bunu göremez. Göremediği için de bir suçu yok, kimse günde on beş forumu aynı anda okuyamıyor.
CTI analysts, we see you. Twelve browser tabs, six forum logins, a Tor session that dies every twenty minutes, and a notes file called "final_final_v3.txt". That's the industry standard workflow for dark web monitoring in 2026, apparently.
So we built the boring solution: one stream. 100+ dark web forums, onion sites included, scraped continuously into a single platform. Roughly a million raw threads so far, growing every hour. No logins, no Tor babysitting, no tab archaeology. You search, you read, you get on with your day.
We're opening researcher access soon. Very few people collect at this scale, and fewer still let anyone else look at it. DMs are open.
samet g. retweeted
🚨Windows Admin Center Remote Code Execution🚨
CVE-2026-56196
CVE-2026-58631
msrc.microsoft.com/update-gu…
msrc.microsoft.com/update-gu…
samet g. retweeted
Our technical write-up on CVE-2026-45504 is now live.
hawktrace.com/blog/CVE-2026-…
#exchange #cve-2026-45504 #hawktrace
samet g. retweeted
Well, well, well. The public JSON formatter sites your developers paste production data into have been quietly publishing every paste for about seven years. Naturally, we read all seven years of it.
200,000+ documents. Cloud keys, SSH keys, payment API keys, whole tax returns with SSNs, people's full identities, bank balances. Nobody hacked anything. People pasted it in to make it look tidy, as you do.
Full writeup below. Yes, it's as bad as it sounds.
İnsanların online JSON formatter'lara yapıştırdığı verileri topladık, ben de Türkiye tarafını epey eşeledim.
Yüzlerce doğrulanmış TCKN. Geçerli IBAN'lar. Birinin araç ruhsatından kart bilgilerine kadar komple kimliği. Hatta bir Türk hosting firmasının kayıtlı müşteri veritabanını bile bulduk, öylece orada duruyordu.
Millet "formatlansın" diye kendi eliyle yapıştırmış.
Well, well, well. The public JSON formatter sites your developers paste production data into have been quietly publishing every paste for about seven years. Naturally, we read all seven years of it.
200,000+ documents. Cloud keys, SSH keys, payment API keys, whole tax returns with SSNs, people's full identities, bank balances. Nobody hacked anything. People pasted it in to make it look tidy, as you do.
Full writeup below. Yes, it's as bad as it sounds.
samet g. retweeted
IbWebAdmin Unauthenticated Remote Code Execution !!!
hawktrace.com/blog/ibwebadmi…
#ibwebadmin #hawktrace #rce #unauth
Introducing samoscout - It combines 53+ native passive sources (more than subfinder, oneforall) with optional active enumeration. It runs subdomain level discovery and even uses a built-in LLM to predict new, undiscovered subdomains.
#recon #bugbounty #cybersecurity #opensource




















