Into The Matrix

Challenge Information

Name: Into The Matrix
Points: 100
Category: AI
Objective: Decode the contents of an NPY file to retrieve the flag.

Solution

Initial Inspection:
- Given an NPY file in CTF, a common format for storing NumPy arrays, the first step is to visualize its contents.

Visualization:

Loaded the NPY file into Python using NumPy:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
 import numpy as np
 import matplotlib.pyplot as plt
 from sklearn.manifold import MDS
 from tqdm import tqdm

 file_path = 'matrix.npy'
 data = np.load(file_path)

 print("Data shape:", data.shape)

 mds = MDS(n_components=2, dissimilarity='precomputed', n_init=4, random_state=42)

 for _ in tqdm(range(2), desc="Performing MDS", total=2):
     pos = mds.fit_transform(data)

 plt.figure(figsize=(10, 10))

 plt.scatter(pos[:, 0], pos[:, 1], c='blue', label='Signal Origins')

 plt.xlabel('MDS Component 1')
 plt.ylabel('MDS Component 2')
 plt.title('MDS Visualization of Distance Matrix')
 plt.legend()

 plt.show()

I wasn’t able to figure out what the graph meant, but my teammate Nem4ros suggested that it appeared to be SHCTF24{intro_to_ml}, but submitting this flag failed.

Brute Forcing Variations:
- Since whe know what the the flag approximately look like, we brute forced the flag, including:
  - SHCTF24{intr0_to_ml}
  - SHCTF24{1ntro_to_ml}
  - SHCTF24{intro_to_ml}
  - …
- Eventually, one of these variations worked as the valid flag.

Flag

Flag: SHCTF24{XXXXXXXXXX}