Base64解码：特定的字符串不正确的填充（具有正确的填充）

Question

I am trying to Base64 decode a string (into bytes) using Python's base64.b64decode(str) method:

46oWrWpy2gTEGwNnN6Ayy

and I am ensuring it has a multiple of 4 ='s for padding or out of frustration any of these:

46oWrWpy2gTEGwNnN6Ayy=

46oWrWpy2gTEGwNnN6Ayy==

46oWrWpy2gTEGwNnN6Ayy===

46oWrWpy2gTEGwNnN6Ayy==================================================

and yet I still get "Incorrect Padding" on Python v3.6.1. Other strings are fine.

I show a colleague, he tries on Python 2 and observes the same response.

I note removing the first "4" is enough to ensure the Base64 decode works.

I have skim read Python's docs (noting casefold doesn't apply for Base64) and haven't yet ventured further into RFC3548 but wondered if someone else had encountered something similar before. Anyone have any clues :)? Surely this can't be a bug in Python's Base64 decoder?

Answer 1

Worked it out.

Each character of Base64 text is 6 bits of the raw's 8 bits. If a character is mid-way through the raw's bytes then you are missing some remaining bits. The Wikipedia article (and many online answers) seems to use padding as interchangeable for a '0' byte which is not the case (in the Base64 dictionary it should be encoded as an A).

Padding is not interchangeable for missing data.

#!/usr/bin/env python3

# We use hexlify for debugging.
import binascii

# We use the Base64 library.
import base64

# Base64 works on multiples of 4 characters..
# ..Sometimes we get 3/2/1 characters and it might be midway through another.
def relaxed_decode_base64(data):

 # If there is already padding we strim it as we calculate padding ourselves.
 if '=' in data:
  data = data[:data.index('=')]

 # We need to add padding, how many bytes are missing.
 missing_padding = len(data) % 4

 # We would be mid-way through a byte.
 if missing_padding == 1:
  data += 'A=='
 # Jut add on the correct length of padding.
 elif missing_padding == 2:
  data += '=='
 elif missing_padding == 3:
  data += '='

 # Actually perform the Base64 decode.
 return base64.b64decode(data)

# Debugging
print(str(relaxed_decode_base64('46oWrWpy2gTEGwNnN6Ayy')) + '\n')

testString = ''

for count in range(0, 1024):
 testString += '/'
 print(str(len(testString)) + ' - ' + testString)
 print(binascii.hexlify(relaxed_decode_base64(testString)))
 input()

Answer 2

Seems to be a problem in your data, not related to Python:

$ echo 46oWrWpy2gTEGwNnN6Ayy | base64 -d
㪭jrÚÄg7 2base64: invalid input
$ echo 46oWrWpy2gTEGwNnN6Ayy= | base64 -d
㪭jrÚÄg7 2base64: invalid input
$ echo 46oWrWpy2gTEGwNnN6Ayy== | base64 -d
㪭jrÚÄg7 2base64: invalid input
$ echo 46oWrWpy2gTEGwNnN6Ayy=== | base64 -d
㪭jrÚÄg7 2base64: invalid input
$ echo 46oWrWpy2gTEGwNnN6Ayy==== | base64 -d
㪭jrÚÄg7 2base64: invalid input

I managed to decode it this way (removed the last 'y'):

$ echo 46oWrWpy2gTEGwNnN6Ay | base64 -d
㪭jrÚÄg7 2