Author:
Leenear

Language:
C/C++

Upload:
2026-03-02 08:35

Platform:
Windows

Difficulty:
4.0

Quality:
4.0

Arch:
x86-64

Downloads:
69

Size:
50.26 KB

Writeups:
0

Comments:
2

Description

A password-protected neural network oracle stored in a custom binary format. Your goal: extract the hidden flag. The model file (model.nli) is an obfuscated binary containing a tiny trained MLP that acts as a password gate. Feed it the right password and it will reveal the flag — but there's no explicit password check in the code. The constraint lives entirely in the network's weights. To solve it you'll need to: reverse the custom .nli binary format, decrypt and parse the model weights, implement inference for a character-level neural network, and search for the password the model was trained to accept. No ML frameworks required — just numpy (or equivalent). The network is small enough to run on CPU instantly. I will share hints every week.

• Comments (2)
• Writeups (0)

You must be logged in to post a comment

You must be logged in to submit a writeup