Dhiraj

Security Researcher, Learning & Fuzzing! (

[Dhiraj @RandomDhiraj]

Windows Terminal SIXEL: when a good prompt met a stubborn OOB write
https://t.co/MRizpaRj86
#infosec #AI

[Dhiraj @RandomDhiraj]

Model quality matters but context is king. While everyone’s hyped about Mythos finding a 27-year-old bug, I tried a similar workflow with current available models.
https://t.co/u6Ldz19adl
#infosec #AI

[Dhiraj @RandomDhiraj]

FFmpeg DFPWM: a 4-year-old integer overflow hiding behind eight samples
https://t.co/wdsK31xO5j
#infosec #AI

[lablab.ai @lablabai]

AI teams and security teams should be finishing each other's sentences.

Join @RandomDhiraj, Security Manager at @Derivdotcom and @czech_pawel, @lablabai Co-Founder & @Surgexyz_ CEO as they share exactly how to make it happen LIVE.

[Deriv People @DerivPeople]

AI teams and security teams should be finishing each other's sentences.

Tune in live with Deriv's Security Manager @RandomDhiraj and @czech_pawel, Co-Founder of @lablabai and CEO of @Surgexyz_, as they share exactly how to make it happen.

They'll get into real breaches, shadow AI, the risks of moving fast, and what tech leaders can do this week to close the gap.

🗓️ Thursday, May 21st · 12:00 PM CET · Live here on this page!

[Dhiraj @RandomDhiraj]

@D0rkerDevil Give a try with SSD-Disclosure it's better than ZDI (Triaging process)

[Dhiraj @RandomDhiraj]

Part-2, Here is the write-up for my recent work RCE in AWS Kiro IDE (CVE-2026–5429).
https://t.co/6d2aUJStXp
#infosec #AI

[Dhiraj @RandomDhiraj]

.@p1ngfl0yd & me would be running 2-day training program on fuzzing at DEFCON this year.
Sign-up for the early bird discount - (https://t.co/rpaLiGzpHa)
#infosec #fuzzing #defcon

[Dhiraj @RandomDhiraj]

Not from AI, just read the existing advisory/patch from VIM (https://t.co/2QNb7u71UA) replaced `scp://` URI to `file://`.

[Dhiraj @RandomDhiraj]

Found command injection via `file://` URI handling in VIM.
Fixed Commit - https://t.co/jYYDwRA9oE
#infosec

[Dhiraj @RandomDhiraj]

Identified a vulnerability in dtprobed (DTrace) where a crafted USDT provider names lead to arbitrary file creation outside intended paths, leading to LPE.
https://t.co/jHaZnNFMI9
#infosec

[Dhiraj @RandomDhiraj]

The red team tradecraft behind hard-to-detect AI phishing.
https://t.co/FPvkXLf6Zd
#infosec #redteam

[Dhiraj @RandomDhiraj]

Here’s my write-up on a code execution in Google Gemini CLI.
https://t.co/CvdNGchwiq
#infosec

[Dhiraj @RandomDhiraj]

I found a uXSS in DuckDuckGo Browser and here is the write-up. https://t.co/Fn28yurQ8I
#infosec

[Dhiraj @RandomDhiraj]

Through libFuzzer, Integer overflow when processing specially crafted ICO image files, leading to memory corruption in GIMP.
https://t.co/7JC1NIqHFA
#infosec #fuzzing

[Dhiraj @RandomDhiraj]

I am sharing it, as this also affects self-hosted runners, usually enterprise servers aren't hardened enough so you can test this during RT/PT. (https://t.co/v6XrWwanyb)

[Dhiraj @RandomDhiraj]

I started reversing GitHub Actions, focusing on escaping the container responsible for running the jobs and found and exploited a volume injection vulnerability that was marked as informative on H1 by GH security.
#infosec #redteam

[Dhiraj @RandomDhiraj]

GH response via H1 - This is an intentional design decision. If you're able to gain a token and access contents outside of your repository's scope, we would certainly be interested.