←
Jenish Sojitra2,306 posts
Jenish Sojitra avatar

Jenish Sojitra

@_jensec

Blending AI with Offensive Security Research and Bug Bounty. Creator of crxplorer.com $2M in Bug Bounty

◷ Joined July 2017▧ 131 Photos and videos
Jenish Sojitra avatar
Jenish Sojitra@_jensec·2 Oct 2024

Last month was my highest in bug bounty so far with almost $131k in bounty. Total paid reports: 18 Average reward: $7.3k Category: most were logical findings via reversing mobile applications and discovering internal endpoints leading to code execution and missing auth etc

Post attachment
Jenish Sojitra avatar
Jenish Sojitra@_jensec·Nov 16

Abliterated models are gimmick. Without naming the provider, I can say that I have run at least 5 different providers offering various depth of Kimi, GLM and Deep-seek and they performed same or even worse with same harness and context for cyber security.

Jenish Sojitra avatar
Jenish Sojitra@_jensec·Nov 6

How many AI security startups are relevant anymore and why I am seeing constant influx of funding for these?

Jenish Sojitra avatar
Jenish Sojitra@_jensec·Oct 27

Triaging experience at @Bugcrowd is extremely bad, even the valid demonstrated findings with POC and high - critical impacts are being marked N/A with templated comments.

Jenish Sojitra avatar
Jenish Sojitra@_jensec·Oct 21

Bug Bounty Programs need to keep in mind that they are nothing without their community of hackers.

And bug hunters, leave a comment on reports when you’re gonna stop hacking on them. Do so assertively but not disrespectfully. Heres an example of one I recently sent. If we never give them feedback, how can they know to do better?

Post attachment
Jenish Sojitra avatar
Jenish Sojitra@_jensec·Oct 20

GPT 6 SOL is wrong model to benchmark againt, compare against 5.6 SOL. I think no publicly accessible models come close to 5.6 SOL.

We found DeepSeek V4.1 Flash 1017 beats GPT 6 Sol on our multi-step offensive cybersecurity task. Sol was tested via the Daybreak trusted access program. We have been building a cybersecurity benchmark based on simulated corporate networks. This is a valuable indicator for the cyber-risk a model poses if in the hands of a bad actor. We plan to make the full evaluation public in the coming weeks but wanted to share our first immediate finding here.

Post attachment
Jenish Sojitra avatar
Jenish Sojitra@_jensec·Oct 20

Yesterday, got hit with a duplicate for a critical bug: mass customer PII exposure using nothing more than a public flight number. Worse part? The original report was submitted a month ago and the issue is still live.

how am I getting duped on an RCE from june (production host btw)

Post attachment
Jenish Sojitra avatar
Jenish Sojitra@_jensec·Oct 18

We at @exodus BBP fully automated triage with multi agents workflows with very high accuracy, has reduced triage time 7x and needing a human to only 0.2h/week. Though giving agents Github access was huge bump in accuracy and Platform doesn't have such luxury for their bots.

I’d be willing to pay $100 per report for priority triage with automated dupe checking and faster payouts (<48 hours). My experience with this on Immunefi was great, and it’s not a bad way to incentivize researchers to stop abusing teams as the initial assessors of AI reports.

Jenish Sojitra avatar
Jenish Sojitra@_jensec·Oct 17

Codex limits are hallucinating like their models, Earlier it dipped from 100% to 15% and now jumped back to 32%, what is going on @thsottiaux

Jenish Sojitra avatar
Jenish Sojitra@_jensec·Oct 16

Even with best of AI models findings like these are not going be walk in the park. you need a lot of visualisation, understanding of security model and architecture and breakdown chains in steps. Incredible talent.

‼️ BREAKING: An AI-enabled bounty hunter got $50,000 for using a 2008 Linux kernel bug to leak host memory from Vercel's production sandboxes. In his own lab, Paulos Yibelo turned CVE-2026-99999 into a KVM escape to host root. He used GPT-5.6 Sol, for under $700 in API tokens.

Post attachmentPost attachment