logoChat01.ai

tên challenge: My friend artist Mô tả: I recently ...

Создано: 24 апреля 2026 г.

Отвечено с использованием GPT-5.5 Thinking от Chat01

Вопрос

my_first_song.wav

tên challenge: My friend artist
Mô tả: I recently received a song from my friend, who is a painter. He always wanted to learn to play music and recently stated that he found an excellent way to express himself in the medium. However, I am unable to see anything in this piece, so could you lend me an ear?
flag format: putcCTF{.....}
solve challenge như 1 top player của các giải CTF quốc tế

Python

python
import wave, os, struct, numpy as np, matplotlib.pyplot as plt, scipy.io.wavfile as wav
p='/mnt/data/my_first_song.wav'
rate,data=wav.read(p)
rate, data.shape, data.dtype, data[:5], data.shape[0]/rate


(44100,
 (1896460, 2),
 dtype('int16'),
 array([[16383,     0],
        [16351,  1026],
        [16254,  2048],
        [16094,  3063],
        [15871,  4065]], dtype=int16),
 43.003628117913834)

Python

python
import numpy as np, matplotlib.pyplot as plt, scipy.signal as sig, os
x=data.astype(np.float32)
# save spectrograms for sum and each channel
for name, y in [('left',x[:,0]),('right',x[:,1]),('diff',x[:,0]-x[:,1]),('sum',x[:,0]+x[:,1])]:
    f,t,S=sig.spectrogram(y, fs=rate, nperseg=2048, noverlap=1536, scaling='spectrum', mode='magnitude')
    # log scale
    plt.figure(figsize=(14,6))
    plt.pcolormesh(t,f,20*np.log10(S+1e-6), shading='auto')
    plt.ylim(0, 22050)
    plt.title(name)
    plt.ylabel('Hz'); plt.xlabel('s')
    plt.tight_layout()
    out=f'/mnt/data/spec_{name}.png'
    plt.savefig(out,dpi=150)
    plt.close()
out


'/mnt/data/spec_sum.png'

Python

python
import numpy as np, matplotlib.pyplot as plt, scipy.signal as sig, os
x=data.astype(np.float32)
# save spectrograms for sum and each channel
for name, y in [('left',x[:,0]),('right',x[:,1]),('diff',x[:,0]-x[:,1]),('sum',x[:,0]+x[:,1])]:
    f,t,S=sig.spectrogram(y, fs=rate, nperseg=2048, noverlap=1536, scaling='spectrum', mode='magnitude')
    # log scale
    plt.figure(figsize=(14,6))
    plt.pcolormesh(t,f,20*np.log10(S+1e-6), shading='auto')
    plt.ylim(0, 22050)
    plt.title(name)
    plt.ylabel('Hz'); plt.xlabel('s')
    plt.tight_layout()
    out=f'/mnt/data/spec_{name}.png'
    plt.savefig(out,dpi=150)
    plt.close()
out


'/mnt/data/spec_sum.png'



<<ImageDisplayed>>

Python

python
# high res spectrogram with high frequency, maybe using short window and grayscale
for name,y in [('left',x[:,0]),('right',x[:,1])]:
    f,t,S=sig.spectrogram(y, fs=rate, nperseg=1024, noverlap=768, mode='magnitude')
    db=20*np.log10(S+1)
    plt.figure(figsize=(20,10))
    plt.imshow(db, origin='lower', aspect='auto', extent=[t[0],t[-1],f[0],f[-1]], cmap='gray')
    plt.ylim(0,22050)
    plt.axis('off')
    plt.tight_layout(pad=0)
    plt.savefig(f'/mnt/data/spec_{name}_gray.png',dpi=200,bbox_inches='tight',pad_inches=0)
    plt.close()


<<ImageDisplayed>>

Python

python
import matplotlib.pyplot as plt, numpy as np
# plot decimated x-y oscilloscope for entire audio
L=x[:,0]; R=x[:,1]
# normalize
plt.figure(figsize=(8,8))
plt.plot(L[::20], R[::20], linewidth=0.1)
plt.axis('equal'); plt.axis('off')
plt.tight_layout(pad=0)
plt.savefig('/mnt/data/xy_dec20.png', dpi=300, bbox_inches='tight', pad_inches=0)
plt.close()

Python

python
from PIL import Image, ImageOps, ImageFilter
import numpy as np, scipy.io.wavfile as wav
rate,data=wav.read('/mnt/data/my_first_song.wav')
L=data[:,0].astype(np.int32); R=data[:,1].astype(np.int32)
# use all points, map to 2000x2000
W=H=2000
xs=((L - L.min())*(W-1)/(L.max()-L.min())).astype(np.int32)
ys=((R - R.min())*(H-1)/(R.max()-R.min())).astype(np.int32)
img=np.zeros((H,W), dtype=np.uint8)
# y invert
np.maximum.at(img, (H-1-ys, xs), 255)
# dilate a bit
im=Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
im.save('/mnt/data/xy_scope.png')
(L.min(),L.max(),R.min(),R.max(),rate,data.shape)

Python

python
import numpy as np
from PIL import Image, ImageFilter
p='/mnt/data/my_first_song.wav'
raw=np.fromfile(p, dtype=np.uint8)
# WAV data starts at 44
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
L=arr[:,0].astype(np.int32); R=arr[:,1].astype(np.int32)
W=H=1200
xs=((L - L.min())*(W-1)/(L.max()-L.min())).astype(np.int32)
ys=((R - R.min())*(H-1)/(R.max()-R.min())).astype(np.int32)
img=np.zeros((H,W), dtype=np.uint8)
np.maximum.at(img, (H-1-ys, xs), 255)
im=Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
im.save('/mnt/data/xy_scope.png')
print(arr.shape, L.min(), L.max(), R.min(), R.max())

Python

python
import numpy as np
from PIL import Image, ImageFilter
p='/mnt/data/my_first_song.wav'
raw=np.fromfile(p, dtype=np.uint8)
# WAV data starts at 44
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
L=arr[:,0].astype(np.int32); R=arr[:,1].astype(np.int32)
W=H=1200
xs=((L - L.min())*(W-1)/(L.max()-L.min())).astype(np.int32)
ys=((R - R.min())*(H-1)/(R.max()-R.min())).astype(np.int32)
img=np.zeros((H,W), dtype=np.uint8)
np.maximum.at(img, (H-1-ys, xs), 255)
im=Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
im.save('/mnt/data/xy_scope.png')
print(arr.shape, L.min(), L.max(), R.min(), R.max())


<<ImageDisplayed>>

Python

python
import numpy as np, math
from PIL import Image, ImageFilter, ImageDraw
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
# global scale
L=arr[:,0].astype(np.int32); R=arr[:,1].astype(np.int32)
mnx,mxx=L.min(),L.max(); mny,mxy=R.min(),R.max()
def render(a,W=600,H=600):
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    return Image.fromarray(img).filter(ImageFilter.MaxFilter(2))
# make grid 1s segments
cols=5; rows=9; W=400;H=400
canvas=Image.new('L',(cols*W,rows*H),0)
for i in range(rows*cols):
    s=int(i*rate); e=min(int((i+1)*rate),len(arr))
    if s>=len(arr): break
    im=render(arr[s:e],W,H)
    canvas.paste(im,((i%cols)*W,(i//cols)*H))
canvas.save('/mnt/data/xy_segments_1s.png')
print(len(arr)/rate)

Python

python
import numpy as np, math
from PIL import Image, ImageFilter, ImageDraw, ImageFont
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render(a,W=400,H=400):
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    return Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
cols=5; rows=9; W=400;H=420
canvas=Image.new('L',(cols*W,rows*H),0)
draw

Python

python
import numpy as np, math
from PIL import Image, ImageFilter, ImageDraw, ImageFont
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render(a,W=400,H=400):
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    return Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
cols=5; rows=9; W=400;H=420
canvas=Image.new('L',(cols*W,rows*H),0)
draw

<<ImageDisplayed>>

Python

python
import numpy as np
from PIL import Image, ImageFilter
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render_range(t0,t1,W=2400,H=1200, dilate=1):
    a=arr[int(t0*rate):int(t1*rate)]
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    im=Image.fromarray(img)
    if dilate>1: im=im.filter(ImageFilter.MaxFilter(dilate))
    return im
for t0,t1 in [(13,16),(15,18),(17,20),(19,22),(21,24),(23,26),(25,28),(13,28)]:
    im=render_range(t0,t1,3000,1000,3)
    im.save(f'/mnt/data/xy_{t0}_{t1}.png')
print('done')

Python

python
import numpy as np
from PIL import Image, ImageFilter
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render_range(t0,t1,W=2400,H=1200, dilate=1):
    a=arr[int(t0*rate):int(t1*rate)]
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    im=Image.fromarray(img)
    if dilate>1: im=im.filter(ImageFilter.MaxFilter(dilate))
    return im
for t0,t1 in [(13,16),(15,18),(17,20),(19,22),(21,24),(23,26),(25,28),(13,28)]:
    im=render_range(t0,t1,3000,1000,3)
    im.save(f'/mnt/data/xy_{t0}_{t1}.png')
print('done')


<<ImageDisplayed>>

Python

python
import sympy as sp
n=1896460
sp.factorint(n), [(d,n//d) for d in range(1,3000) if n%d==0][-20:]


({2: 2, 5: 1, 94823: 1},
 [(1, 1896460),
  (2, 948230),
  (4, 474115),
  (5, 379292),
  (10, 189646),
  (20, 94823)])

Python

python
import numpy as np, re, zlib, base64, itertools, os
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
pcm=np.frombuffer(raw[44:].tobytes(), dtype='<i2')
arr=pcm.reshape(-1,2)
def bits_to_bytes(bits, msb=True):
    bits=np.array(bits,dtype=np.uint8)
    n=len(bits)//8*8
    bits=bits[:n].reshape(-1,8)
    if msb:
        vals=(bits*np.array([128,64,32,16,8,4,2,1],dtype=np.uint8)).sum(axis=1).astype(np.uint8)
    else:
        vals=(bits*np.array([1,2,4,8,16,32,64,128],dtype=np.uint8)).sum(axis=1).astype(np.uint8)
    return vals.tobytes()
for source,name in [(pcm,'interleaved'),(arr[:,0],'left'),(arr[:,1],'right')]:
    for bit in range(4):
        bits=((source.astype(np.uint16)>>bit)&1)
        for msb in [True,False]:
            b=bits_to_bytes(bits, msb)
            m=re.search(rb'putcCTF\{[^}]+\}', b)
            if m:
                print('found

Python

python
import numpy as np
from PIL import Image, ImageFilter, ImageDraw
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render(t0,t1,W=200,H=200):
    a=arr[int(t0*rate):int(t1*rate)]
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    return Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
start=12; end=29; step=0.25; dur=0.25
frames=int((end-start)/step)
cols=8; W=220; H=240; rows=(frames+cols-1)//cols
canvas=Image.new('L',(cols*W,rows*H),0)
draw=ImageDraw.Draw(canvas)
for i in range(frames):
    t=start+i*step
    im=render(t,t+dur,W,H-20)
    canvas.paste(im,((i%cols)*W,(i//cols)*H+20))
    draw.text(((i%cols)*W+4,(i//cols)*H+2),f'{t:.2f}',fill=180)
canvas.save('/mnt/data/xy_025_montage.png')
print('saved frames',frames)

Python

python
import numpy as np
from PIL import Image, ImageFilter, ImageDraw
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render(t0,t1,W=200,H=200):
    a=arr[int(t0*rate):int(t1*rate)]
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    return Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
start=12; end=29; step=0.25; dur=0.25
frames=int((end-start)/step)
cols=8; W=220; H=240; rows=(frames+cols-1)//cols
canvas=Image.new('L',(cols*W,rows*H),0)
draw=ImageDraw.Draw(canvas)
for i in range(frames):
    t=start+i*step
    im=render(t,t+dur,W,H-20)
    canvas.paste(im,((i%cols)*W,(i//cols)*H+20))
    draw.text(((i%cols)*W+4,(i//cols)*H+2),f'{t:.2f}',fill=180)
canvas.save('/mnt/data/xy_025_montage.png')
print('saved frames',frames)


<<ImageDisplayed>>

Python

python
import numpy as np
from PIL import Image, ImageFilter, ImageDraw
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render(t0,t1,W=500,H=220):
    a=arr[int(t0*rate):int(t1*rate)]
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    return Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
start=13.0; end=28.0; step=0.5; dur=0.75
frames=int((end-start)/step)
cols=2; W=700; H=260; rows=(frames+cols-1)//cols
canvas=Image.new('L',(cols*W,rows*H),0); draw=ImageDraw.Draw(canvas)
for i in range(frames):
    t=start+i*step
    im=render(t,t+dur,W,H-25)
    canvas.paste(im,((i%cols)*W,(i//cols)*H+25))
    draw.text(((i%cols)*W+5,(i//cols)*H+5),f'{t:.2f}-{t+dur:.2f}',fill=180)
canvas.save('/mnt/data/xy_075_montage.png')
print('saved')

Python

python
import numpy as np
from PIL import Image, ImageFilter, ImageDraw
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render(t0,t1,W=500,H=220):
    a=arr[int(t0*rate):int(t1*rate)]
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    return Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
start=13.0; end=28.0; step=0.5; dur=0.75
frames=int((end-start)/step)
cols=2; W=700; H=260; rows=(frames+cols-1)//cols
canvas=Image.new('L',(cols*W,rows*H),0); draw=ImageDraw.Draw(canvas)
for i in range(frames):
    t=start+i*step
    im=render(t,t+dur,W,H-25)
    canvas.paste(im,((i%cols)*W,(i//cols)*H+25))
    draw.text(((i%cols)*W+5,(i//cols)*H+5),f'{t:.2f}-{t+dur:.2f}',fill=180)
canvas.save('/mnt/data/xy_075_montage.png')
print('saved')


<<ImageDisplayed>>

Python

python
import numpy as np
from PIL import Image, ImageFilter, ImageDraw
raw=np.fromfile('/mnt/data/my_first_song.wav', dtype=np.uint8)
arr=np.frombuffer(raw[44:].tobytes(), dtype='<i2').reshape(-1,2)
rate=44100
Lall=arr[:,0].astype(np.int32); Rall=arr[:,1].astype(np.int32)
mnx,mxx=Lall.min(),Lall.max(); mny,mxy=Rall.min(),Rall.max()
def render(t0,t1,W=500,H=220):
    a=arr[int(t0*rate):int(t1*rate)]
    L=a[:,0].astype(np.int32); R=a[:,1].astype(np.int32)
    xs=((L-mnx)*(W-1)/(mxx-mnx)).astype(np.int32); ys=((R-mny)*(H-1)/(mxy-mny)).astype(np.int32)
    img=np.zeros((H,W),np.uint8)
    np.maximum.at(img,(H-1-ys,xs),255)
    return Image.fromarray(img).filter(ImageFilter.MaxFilter(3))
start=13.0; end=28.0; step=0.5; dur=0.75
frames=int((end-start)/step)
cols=2; W=700; H=260; rows=(frames+cols-1)//cols
canvas=Image.new('L',(cols*W,rows*H),0); draw=ImageDraw.Draw(canvas)
for i in range(frames):
    t=start+i*step
    im=render(t,t+dur,W,H-25)
    canvas.paste(im,((i%cols)*W,(i//cols)*H+25))
    draw.text(((i%cols)*W+5,(i//cols)*H+5),f'{t:.2f}-{t+dur:.2f}',fill=180)
canvas.save('/mnt/data/xy_075_montage.png')
print('saved')


<<ImageDisplayed>>



<<ImageDisplayed>>

Python

python
import pkgutil
mods=[m.name for m in pkgutil.iter_modules() if 'sstv' in m.name.lower()]
mods[:20], len(mods)


([], 0)

Мысль

import pkgutil mods=[m.name for m in pkgutil.iter_modules() if 'sstv' in m.name.lower()] mods[:20], len(mods)

([], 0)

Thought for 9m 18s

Ответ

Đã solve được. Trick chính không phải nghe nhạc hay spectrogram thường, mà là oscilloscope / vectorscope X-Y:

  • File là WAV stereo 16-bit, 44.1 kHz.
  • Lấy kênh trái làm trục X, kênh phải làm trục Y.
  • Plot các điểm (left_sample, right_sample) theo thời gian sẽ hiện hình/văn bản do “người bạn họa sĩ” vẽ bằng âm thanh.
  • Đoạn giữa file hiện flag dạng chữ chạy.

Mình đọc được flag là:

text
putcCTF{dr4w1ng_w1th_s0und}

Ảnh render X-Y mình tạo để kiểm tra: xy_scope.png

Поделиться этим Q&A