Since I'm not planning on re-doing a live session of my Mastering Binary Files & Protocols workshop any time soon, I've made the recordings from the last session available in our web shop:
hackarcana.com/workshop-sess…
In terms of the pricing:
1. Price has been cut in half: 600 CHF net (+tax if applicable) from previous 1350 CHF net (+tax)
2. The workshop has been split into individual modules, so you don't have to get e.g. the basics if you don't need them, or the advanced stuff if you're not sure about it. Prices for modules are: 75, 150, 150, 150, 200 CHF net (+tax if applicable)
3. Furthermore, it's Black Friday season, so here's another 15% off with this code: MBF-BLACK-FRIDAY (until end of month)
That's 43 hours of recordings + 33 CTF-like exercises that can get you from "I don't know anything about binary file/network data formats" to "format specs are find, but if there aren't any, I still know how to deal with it" ;)
Enjoy!
I'm reading Marcus' recent post and I'm not exactly sure where is this coming from. A few things:
1. FFmpeg has a history of quickly fixing reported bugs, and before 2025 Google has reported literally thousands of security vulnerabilities to FFmpeg, which were fixed within days. Then FFmpeg switched to a new X/Twitter PR approach and suddenly reporting all discovered bugs somehow became bad?
2. FFmpeg is known to be broken. It doesn't matter if the bugs are disclosed after 90 days, 60 days, 30 days, or immediately, as running FFmpeg outside of the sandbox on untrusted input has been known to be a high-risk approach for way over a decade if not two. It's just to easy to find vulnerabilities in it and too hard to get stuff right in languages like C or ASM.
3. What's the exact alternative being advocated for here? Not reporting all the bugs because too many were found? Having people unfamiliar with the whole code base make patches (all 2 of them, or however many the exact research group that found the bugs had)? Having AI make patches? Or maybe keeping users in the dark and delaying telling them about security issues in the software they use (see next paragraph)?
If a software project has security vulnerabilities, the approach is to inform the owners as soon as possible and - more importantly - the users within reasonable time, so they - if they wish to do so and are interested in doing so - are able to take precautions and apply workarounds. Arguing against N-day disclosure policies is arguing against allowing users to make informed choices wrt the software they run. And IMO 90 days is already too long for users to be kept in the dark - especially in 2026.
Overall I don't think the FFmpeg crash out was valid.
I am still happy that the FFmpeg project received some funding thanks to the whole drama though, and I would wish for it to be receiving even more from its large commercial users (but that's a different discussion and not achievable without a - probably impossible - switch to dual licensing).
🎟️ Get 15% off either workshop set with code MBF-BLACK-FRIDAY until November 30!
The discount applies to both individual modules and sets.
👉 hackarcana.com/mbf
Working on frontier cybersecurity evals & benchmarks be like 0days as environments. Found a guest to host escape with host-user code execution in the latest VirtualBox version (7.2.22). Full write-up after coordinated disclosure with Oracle is complete.
AI-driven information operations will reshape security and will be a challenge to societies. I'll speak about this at a seminar at @KingsCollegeLon@warstudies on 10 December/13:00. I'll show the first fully autonomous information operation. AI agents running influence campaigns end to end. What if AI starts selecting campaign objectives too?
To expand on this a bit: Yes, the LLM found more bugs than the fuzzer. But the fuzzer finding bugs the LLM didn't shouldn't come as a surprise to anyone who has ever performed a thorough manual review, only to run a fuzzer afterward and discover issues they missed.
Universal exploit for #php2shell is rather impossible, but specialized exploits for known environments are coming. I tried my luck with pfSense 2.7.2 - PHP 8.2.11, FPM, 512M limit. Variant with session file write to a crafted path. Almost entirely written by an open-weights LLM.
With all these exploits flying around on the internet I think we should all switch to HTTP 4.0, which I would like to propose here:
HTTP 4.0: Human hand-To-hand Transfer Protocol
A post-apo protocol where you just hand your friend your website on a pendrive/CD/floppy/tape.
You MUST first check if they are not a replicant of course.
There are a technical lot of things I will miss that AI basically made obsolete.
But one thing I won't miss is DEBUGGING RANDOM ISSUES ON MY LINUX DESKTOP! AI deals with them super fast - basically a solved problem. And good riddance.
Year of Linux on desktop?