| dc540-0x0000a |
Appreciate the honest reply, and I take your point about recon and realizing what data is missing.
But I think moving the secret out of the binary works against what a crackme actually is. The essence of a crackme is that the answer lives in the artifact and you earn it by understanding the code. When the key material is posted somewhere online instead, the task stops being reverse engineering and turns into find-where-the-author-dropped-it. That is OSINT at best, and honestly closer to guessing where you posted your secrets. I should not have to comb social feeds to finish a binary.
In CTF terms that is a guess, and guessy is treated as a design smell for a good reason. It does break the instant-LLM loop, but it breaks it by removing the thing that makes the challenge a crackme, not by making the reversing harder. A self-contained binary that is genuinely hard to reverse would stop the automated solvers and still be a real crackme. This one stops them by not being fully solvable from the download.
Still a cool build and a fun teardown. I just would not call the withheld-key part good crackme design.
|
2026-11-15 16:38 |
| dc540-0x0000a |
Really enjoyed tearing into this one. The bare-metal RP2350 build is clean and the AEAD-sealed record table is a nice piece of work.
But I want to be straight about what it is. The firmware reverses fully. The dispatcher, the sealed records, the KDF, and Gate 1 all give themselves up, and Gate 1 authenticates end to end with the Ash Word you posted. Everything past that is the problem. The fragments that complete the flag are not in the image. The binary only carries the gate that validates a word, never the word itself, and each one is a 256-bit key behind an AEAD tag, so it cannot be derived or guessed. They exist only in your social drops, on your timer.
So this is not solvable from the artifact alone, by a person or a model. That makes it a timed ARG where the download is one piece of several, not a self-contained crackme. Which is fine as an ARG, but it is worth being honest that "AI cannot solve it" is not really about the reverse engineering. The RE is done. The rest is just waiting for the next word to get posted.
Curious how you see it.
|
2026-11-13 10:44 |