Found a Medium-severity vulnerability in Zebra, @zcash's Rust full-node implementation, and reported it through the Zcash Bounty Program for a $37,500 bounty. Already patched and shipped in v4.5.0 — disclosure below. github.com/ZcashFoundation/z… @ZcashCommGrants Had some duplicates too. Feels good 😁

Been grinding hard lately. Over 80 private audits and bug bounties across different protocols. Honestly, it’s been exhausting. Along the way I’ve been using Chinese models a lot more for the work especially GLM 5.3. I stopped relying on Claude because of the guardrails. Hard to get real hands on the latest models (Opus 5.5 and Mythos) for actual security research. GLM doesn’t have those heavy restrictions. It’s more instruction-specific though. With Claude you can throw something abstract and it usually figures out the intent. With GLM you need precise, specific instructions but when you give it those, it executes cleanly and gets results. I’ve been running it on a bunch of protocols and pulling solid findings. For security research it’s as good as any other model I’ve used. Don’t know if every benchmark floating around on X is accurate, but in practice it’s a very good tool for this. The real frustration right now though? Private bug bounty programs. No timelines. Protocols can leave you waiting 1-3 months with zero response. No clear rules. Nobody to answer to. By the time you hear back on a valid finding you’re already exhausted and sometimes you’ve half-forgotten the protocol. It’s slow. It’s draining. But we keep pushing. Still hunting for bigger payouts. Still aiming for something substantial one day maybe even that $1M mark. Anyone else dealing with the private program lag and the volume at the same time? How are you handling it? Still hunting. Still learning.

These are all the facts you need

𒀸◦✧◈ 𝐀𝐧𝐚𝐭𝐨𝐦𝐲 𝐨𝐟 𝐚 𝐫𝐞𝐦𝐨𝐭𝐞 𝐃𝐨𝐒 ◈✧◦𒇷 a malformed typed input slips past every pre-check and hits an unguarded stack[0] read in core validation → SIGSEGV. The guard that would've rejected it sits downstream but never reached. Full write-up after coordinated disclosure :)

I personally haven’t said anything about this because it’s been nothing but uttermost pain seeing how we whitehats are being treated with not even a single recognition or even as simple as “Thank you” sometimes from the protocols. Since the overwhelming increase in the number of security researchers participating in bug bounties and audit competitions on bug bounty platforms (@immunefi) leading to the rise in duplicates and undefined closures of vulnerabilities reported, I then decided to fully move into Private Audits focusing on only Critical/High severity vulnerabilities on all blockchains on @DefiLlama and over the past few months. I have audited >60 chains and found multiple C/H findings but sadly I have only been able to communicate directly with less than 5 chains of which they are still even here to give me a first response or acknowledgement that they have even seen my report and are looking into it which is so disappointing. I honestly don’t know what to say again as i am slowly giving up the discipline of reporting directly to their teams but instead now want to be posting my findings publicly on here as an article hopefully it will draw their attention. @Meter_IO @VictionEco @iota and so much more out there…. BE WARNED ‼️

Six teams used FROST, but they did not build the same thing. I compared their custody models, thresholds, demos, network evidence, recovery paths, licenses and current limitations—using only what each submission and repository documents.

All thanks to @yassine3eth I got my account back on @immunefi…. Learnt a lot in the dark days, back now stronger and 10x better. Watch out :)

Reported a medium-severity remote DoS vulnerability in @PirateChain capable of crashing synchronized production nodes. Rooted in the shared Komodo codebase, so it affected the whole chain family. Responsibly disclosed, coordinated, and now fixed in v5.9.3 piratechain.com/blog/pirate-…

𝕋𝕙𝕖 𝔹𝕖𝕤𝕥 𝕊𝕥𝕣𝕒𝕥𝕖𝕘𝕪

Very true!! But now I find the concept of private bug bounties better as most times when protocols in general hosts bug bounty program on a web3 bug bounty platform, they end up not fully maximizing the use of it because of the number of spam submissions/AI slop that eventually end up getting the protocol team frustrated and then pausing their bounty program

Large Language Models > Self training :))

Just keep auditing and verifying that AI slop till you get a response from your LLM saying: "If you don't believe me then forget"

Congratulations brother 😁👏 We keep winning!!!