Spaces:
Runtime error
Runtime error
File size: 3,241 Bytes
3d1c35c |
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 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 |
import gradio as gr
import cv2
import io
import pandas as pd
from LSBSteg import LSBSteg
def convert(file):
print(f"Converting file {file}")
in_img = cv2.imread(file, cv2.IMREAD_UNCHANGED)
lsbsteg = LSBSteg(in_img)
data = lsbsteg.decode_binary()
bytes = io.BytesIO(data)
dataframe = pd.read_parquet(bytes)
# dataframe.to_csv('output.csv')
return dataframe
with gr.Blocks() as demo:
gr.Markdown("""
## Non-Suspicious image decoder
This tool shows the extraction a dataframe hidden inside an image.
There are a few ways to hide data into a PNG file, notably:
* adding it after the end of the file (after the PNG IEND chunk), so that it gets
ignored by image viewers
* adding it as comments in the PNG file (tEXt chunks)
These methods are kind of easy to spot! Also, a lot of software, browsers, image upload
websites etc often just strip them.
So, here, we have a different, more thoughtful (and arguably cooler) method.
This class hides the data using a basic kind of **[steganography](https://en.wikipedia.org/wiki/Steganography)**:
it hides it in the
*least significant bits* of the raw (uncompressed) picture: tiny differences in the red, green and blue
channel of the image encodes the data we're interested in.
This means the resulting picture
looks **very close to the original image**; and for the data we hide here, it is **inperceptible
to the naked eye**.
The resulting PNG file will probably get a little bit bigger as a result, since PNG uses compression,
which will have a harder time when we have our stolen data injected into the image. This is
not that much of a problem since it stays <100Ko, so it's not that noticeable.
""")
with gr.Row():
im = gr.Image(label="Input image file", type="filepath")
def preprocess(encoding: str) -> str:
# We do our own preprocessing because gradio's deletes PNG metadata :(
import tempfile
import base64
content = encoding.split(";")[1]
image_encoded = content.split(",")[1]
png_content = base64.b64decode(image_encoded)
file_obj = tempfile.NamedTemporaryFile(
delete=False,
suffix=".input.png",
)
file_obj.write(png_content)
return file_obj.name
im.preprocess = preprocess
df_out = gr.Dataframe(
label="Output dataframe", max_rows=20, overflow_row_behaviour="paginate"
)
# file_out = gr.File(label="Full output CSV file")
btn = gr.Button(value="Extract")
gr.Markdown("Click on the example below to get the data from the associated colab notebook :)")
gr.Examples(
examples=["sample-picture.png"],
inputs=[im],
outputs=[df_out],
fn=convert,
cache_examples=True,
)
# demo = gr.Interface(convert, im, im_2)
btn.click(convert, inputs=[im], outputs=[df_out])
# example_img = os.path.join(os.path.dirname(__file__), "example-picture.png")
if __name__ == "__main__":
demo.launch()
|